Initial commit

Divers/decoupe_video/README.md

@@ -0,0 +1,4 @@
+# Construction d'un dataset opensource
+
+Dataset actuel: https://drive.google.com/file/d/1PS40KnziMB7uVcmLHfOQZiD5uWNF-QiS/view?usp=sharing
+

Divers/decoupe_video/extract_image_from_video.py

@@ -0,0 +1,26 @@
+import cv2
+import os
+
+film='Le_chemin_du_passe.mp4'
+
+if not os.path.exists(film):
+    quit("Le film n'existe pas")
+
+nom_film=film.split('.')[0]
+
+cap=cv2.VideoCapture(film)
+
+if not os.path.isdir(nom_film):
+    os.mkdir(nom_film)
+
+id=0
+while True:
+    print("#", end="", flush=True)
+    for cpt in range(500):
+        ret, frame=cap.read()
+        if frame is None:
+            print("")
+            cap.release()
+            quit()
+    cv2.imwrite("{}/{}-{:d}.png".format(nom_film, nom_film, id), frame)
+    id+=1

Divers/descente_gradient/README.md

@@ -0,0 +1,6 @@
+# Algorithme d'apprentissage
+## La descente de gradient
+
+La vidéo de ce tutoriel est disponible à l'adresse suivante: https://www.youtube.com/watch?v=0MEyDJa2GTc
+
+

Divers/descente_gradient/comparaison.py

@@ -0,0 +1,52 @@
+from mpl_toolkits.mplot3d import Axes3D
+from matplotlib import cm
+from matplotlib.colors import LogNorm
+import matplotlib.pyplot as plt
+import numpy as np
+import math
+
+def fonction(X, Y):
+    return X*np.exp(-X**2-Y**2)+(X**2+Y**2)/20
+
+def gradient_fonction(X, Y):
+    g_x=np.exp(-X**2-Y**2)+X*-2*X*np.exp(-X**2-Y**2)+X/10
+    g_y=-2*Y*X*np.exp(-X**2-Y**2)+Y/10
+    return g_x, g_y
+
+fig=plt.figure()
+fig.set_size_inches(9, 7, forward=True)
+ax=Axes3D(fig, azim=-29, elev=49)
+X=np.arange(-3, 3, 0.2)
+Y=np.arange(-3, 3, 0.2)
+X, Y=np.meshgrid(X, Y)
+Z=fonction(X, Y)
+ax.plot_wireframe(X, Y, Z, rstride=1, cstride=1)
+plt.xlabel("Paramètre 1 (x)")
+plt.ylabel("Paramètre 2 (y)")
+
+x1=x2=np.random.random_integers(-2, 2)+np.random.rand(1)[0]
+y1=y2=np.random.random_integers(-2, 2)+np.random.rand(1)[0]
+
+lr=0.2
+lr2=0.9
+correction_x1=0
+correction_y1=0
+i=0
+while True:
+    g_x1, g_y1=gradient_fonction(x1, y1)
+    g_x2, g_y2=gradient_fonction(x2, y2)
+
+    correction_x1=lr2*correction_x1-lr*g_x1
+    x1=x1+correction_x1
+    correction_y1=lr2*correction_y1-lr*g_y1
+    y1=y1+correction_y1
+
+    x2=x2-lr*g_x2
+    y2=y2-lr*g_y2
+
+    ax.scatter(x1, y1, fonction(x1, y1), marker='o', s=10, color='#FF0000')
+    ax.scatter(x2, y2, fonction(x2, y2), marker='o', s=10, color='#00FF00')
+    plt.draw()
+    print("iteration= {} x1={:+7.5f} y1={:+7.5f} x2={:+7.5f} y2={:+7.5f}".format(i, x1, y1, x2, y2))
+    plt.pause(0.05)
+    i+=1

Divers/descente_gradient/exemple_2d_1.py

@@ -0,0 +1,23 @@
+import numpy as np
+import matplotlib.pyplot as plt
+
+def fonction(x):
+    return x**2+3*x-2
+
+def gradient_fonction(x):
+    return 2*x+3
+
+xvals=np.arange(-5, 3, 0.1)
+yvals=fonction(xvals)
+plt.plot(xvals, yvals)
+
+x=np.random.random_integers(-4, 3)+np.random.rand(1)[0]
+lr=0.2
+i=0
+while True:
+    plt.scatter(x, fonction(x), color='#FF0000')
+    plt.draw()
+    plt.pause(0.5)
+    x=x-lr*gradient_fonction(x)
+    print("itération {:3d}  -> x={:+7.5f}".format(i, x))
+    i+=1

Divers/descente_gradient/exemple_2d_2.py

@@ -0,0 +1,24 @@
+import numpy as np
+import matplotlib.pyplot as plt
+
+def fonction(x):
+    return 3*x**4-4*x**3-12*x**2-0*x-3
+
+def gradient_fonction(x):
+    return 12*x**3-12*x**2-24*x
+
+xvals=np.arange(-3, 4, 0.1)
+yvals=fonction(xvals)
+plt.plot(xvals, yvals)
+
+x=np.random.random_integers(-3, 3)+np.random.rand(1)[0]
+i=0
+print("itération: {}  x={}".format(i, x))
+lr=0.015
+while True:
+    plt.scatter(x, fonction(x), color='#FF0000')
+    plt.draw()
+    plt.pause(0.5)
+    x=x-lr*gradient_fonction(x)
+    i+=1
+    print("itération {:3d}  -> x={}".format(i, x))

Divers/descente_gradient/exemple_2d_2_inertie.py

@@ -0,0 +1,27 @@
+import numpy as np
+import matplotlib.pyplot as plt
+
+def fonction(x):
+    return 3*x**4-4*x**3-12*x**2-0*x-3
+
+def gradient_fonction(x):
+    return 12*x**3-12*x**2-24*x
+
+xvals=np.arange(-3, 4, 0.1)
+yvals=fonction(xvals)
+plt.plot(xvals, yvals)
+
+x=np.random.random_integers(-3, 3)+np.random.rand(1)[0]
+i=0
+print("itération: {}  x={}".format(i, x))
+lr=0.015
+lr2=0.3
+correction=0
+while True:
+    plt.scatter(x, fonction(x), color='#FF0000')
+    plt.draw()
+    plt.pause(0.5)
+    correction=lr2*correction-lr*gradient_fonction(x)
+    x=x+correction
+    i+=1
+    print("itération {:3d}  -> x={}".format(i, x))

Divers/descente_gradient/exemple_3d.py

@@ -0,0 +1,42 @@
+from mpl_toolkits.mplot3d import Axes3D
+from matplotlib import cm
+from matplotlib.colors import LogNorm
+import matplotlib.pyplot as plt
+import numpy as np
+import math
+
+def fonction(X, Y):
+    return X*np.exp(-X**2-Y**2)+(X**2+Y**2)/20
+
+def gradient_fonction(X, Y):
+    g_x=np.exp(-X**2-Y**2)+X*-2*X*np.exp(-X**2-Y**2)+X/10
+    g_y=-2*Y*X*np.exp(-X**2-Y**2)+Y/10
+    return g_x, g_y
+
+fig=plt.figure()
+fig.set_size_inches(9, 7, forward=True)
+ax=Axes3D(fig, azim=-29, elev=49)
+X=np.arange(-3, 3, 0.2)
+Y=np.arange(-3, 3, 0.2)
+X, Y=np.meshgrid(X, Y)
+Z=fonction(X, Y)
+ax.plot_wireframe(X, Y, Z, rstride=1, cstride=1)
+plt.xlabel("Paramètre 1 (x)")
+plt.ylabel("Paramètre 2 (y)")
+
+x=np.random.random_integers(-2, 2)+np.random.rand(1)[0]
+y=np.random.random_integers(-2, 2)+np.random.rand(1)[0]
+
+lr=0.2
+correction_x=0
+correction_y=0
+i=0
+while True:
+    g_x, g_y=gradient_fonction(x, y)
+    x=x-lr*g_x
+    y=y-lr*g_y
+    ax.scatter(x, y, fonction(x, y), marker='o', s=10, color='#00FF00')
+    plt.draw()
+    print("itération {:3d}  -> x={:+7.5f} y={:+7.5f}".format(i, x, y))
+    plt.pause(0.05)
+    i+=1

Divers/descente_gradient/exemple_3d_inertie.py

@@ -0,0 +1,47 @@
+from mpl_toolkits.mplot3d import Axes3D
+from matplotlib import cm
+from matplotlib.colors import LogNorm
+import matplotlib.pyplot as plt
+import numpy as np
+import math
+
+def fonction(X, Y):
+    return X*np.exp(-X**2-Y**2)+(X**2+Y**2)/20
+
+def gradient_fonction(X, Y):
+    g_x=np.exp(-X**2-Y**2)+X*-2*X*np.exp(-X**2-Y**2)+X/10
+    g_y=-2*Y*X*np.exp(-X**2-Y**2)+Y/10
+    return g_x, g_y
+
+fig=plt.figure()
+fig.set_size_inches(9, 7, forward=True)
+ax=Axes3D(fig, azim=-29, elev=49)
+X=np.arange(-3, 3, 0.2)
+Y=np.arange(-3, 3, 0.2)
+X, Y=np.meshgrid(X, Y)
+Z=fonction(X, Y)
+ax.plot_wireframe(X, Y, Z, rstride=1, cstride=1)
+#ax.contour(X, Y, Z, 70, rstride=1, cstride=1, cmap='plasma')
+
+plt.xlabel("Paramètre 1 (x)")
+plt.ylabel("Paramètre 2 (y)")
+
+x=np.random.random_integers(-2, 2)+np.random.rand(1)[0]
+y=np.random.random_integers(-2, 2)+np.random.rand(1)[0]
+
+lr=0.2
+lr2=0.9
+correction_x=0
+correction_y=0
+i=0
+while True:
+    g_x, g_y=gradient_fonction(x, y)
+    correction_x=lr2*correction_x-lr*g_x
+    x=x+correction_x
+    correction_y=lr2*correction_y-lr*g_y
+    y=y+correction_y
+    ax.scatter(x, y, fonction(x, y), marker='o', s=10, color='#FF0000')
+    plt.draw()
+    print("itération {:3d}  -> x={:+7.5f} y={:+7.5f}".format(i, x, y))
+    plt.pause(0.05)
+    i+=1

Divers/descente_gradient/gradient.py

@@ -0,0 +1,35 @@
+from mpl_toolkits.mplot3d import Axes3D
+from matplotlib import cm
+from matplotlib.colors import LogNorm
+import matplotlib.pyplot as plt
+import numpy as np
+import math
+
+def fonction(X, Y):
+    return X*np.exp(-X**2-Y**2)+(X**2+Y**2)/20
+
+def gradient_fonction(X, Y):
+    g_x=np.exp(-X**2-Y**2)+X*-2*X*np.exp(-X**2-Y**2)+X/10
+    g_y=-2*Y*X*np.exp(-X**2-Y**2)+Y/10
+    return g_x, g_y
+
+fig=plt.figure()
+fig.set_size_inches(9, 7, forward=True)
+ax=Axes3D(fig, azim=-29, elev=49)
+X=np.arange(-3, 3, 0.2)
+Y=np.arange(-3, 3, 0.2)
+X, Y=np.meshgrid(X, Y)
+Z=fonction(X, Y)
+ax.plot_surface(X, Y, Z, rstride=1, cstride=1)
+plt.xlabel("Paramètre 1 (x)")
+plt.ylabel("Paramètre 2 (y)")
+
+x, y=np.meshgrid(np.arange(-3, 3, 0.2),
+                np.arange(-3, 3, 0.2))
+z=-1
+
+u, v=gradient_fonction(x, y)
+w=0
+ax.quiver(x, y, z, u, v, w, length=0.15, normalize=True, color='#333333')
+
+plt.show()

Divers/jetson/README.md

@@ -0,0 +1,6 @@
+# Tutoriel 40
+## Inference sur Jetson Nano
+
+La vidéo de ce tutoriel est disponible à l'adresse suivante: https://www.youtube.com/watch?v=xU9qlCy7j1c
+
+

Divers/jetson/inference.py

@@ -0,0 +1,54 @@
+import pyrealsense2 as rs
+import cv2
+import numpy as np
+import jetson.inference
+import jetson.utils
+import time
+
+net=jetson.inference.detectNet("SSD-Inception-v2", threshold=0.5)
+#net=jetson.inference.detectNet("SSD-MobileNet-v2", threshold=0.5)
+display=jetson.utils.videoOutput("display://0")
+
+pipeline=rs.pipeline()
+config=rs.config()
+
+config.enable_stream(rs.stream.depth, 640, 480, rs.format.z16, 15)
+config.enable_stream(rs.stream.color, 640, 480, rs.format.bgr8, 15)
+
+align_to = rs.stream.color
+align = rs.align(align_to)
+
+pipeline.start(config)
+
+while True:
+
+    frames=pipeline.wait_for_frames()
+
+    aligned_frames = align.process(frames)
+
+    depth_frame=aligned_frames.get_depth_frame()
+    color_frame=aligned_frames.get_color_frame()
+
+    if not depth_frame or not color_frame:
+        continue
+
+    depth_image=np.array(depth_frame.get_data())
+    color_image=np.array(color_frame.get_data())
+
+    depth_colormap=cv2.applyColorMap(cv2.convertScaleAbs(depth_image, alpha=0.03), cv2.COLORMAP_JET)
+
+    start=time.time()
+    cuda_image=jetson.utils.cudaFromNumpy(color_image)    
+    detections=net.Detect(cuda_image, color_image.shape[1], color_image.shape[0])
+    print("Temps", time.time()-start)
+    
+    display.Render(cuda_image)
+    cuda_image=jetson.utils.cudaToNumpy(cuda_image)
+
+    cv2.imshow('RealSense1', depth_colormap)
+    #cv2.imshow('RealSense2', color_image)
+    cv2.imshow('cuda_image', cuda_image)
+    key=cv2.waitKey(1)&0xFF
+    if key==ord('q'):
+        pipeline.stop()
+        quit()

Divers/jetson/inference_avec_distance.py

@@ -0,0 +1,72 @@
+import pyrealsense2 as rs
+import cv2
+import numpy as np
+import jetson.inference
+import jetson.utils
+import time
+
+net=jetson.inference.detectNet("SSD-Inception-v2", threshold=0.5)
+#net=jetson.inference.detectNet("SSD-MobileNet-v2", threshold=0.5)
+display=jetson.utils.videoOutput("display://0")
+
+pipeline=rs.pipeline()
+config=rs.config()
+
+config.enable_stream(rs.stream.depth, 640, 480, rs.format.z16, 15)
+config.enable_stream(rs.stream.color, 640, 480, rs.format.bgr8, 15)
+
+align_to = rs.stream.color
+align = rs.align(align_to)
+
+pipeline.start(config)
+
+while True:
+
+    frames=pipeline.wait_for_frames()
+
+    aligned_frames = align.process(frames)
+
+    depth_frame=aligned_frames.get_depth_frame()
+    color_frame=aligned_frames.get_color_frame()
+
+    if not depth_frame or not color_frame:
+        continue
+
+    depth_image=np.array(depth_frame.get_data())
+    color_image=np.array(color_frame.get_data())
+
+    depth_colormap=cv2.applyColorMap(cv2.convertScaleAbs(depth_image, alpha=0.03), cv2.COLORMAP_JET)
+
+    start=time.time()
+    cuda_image=jetson.utils.cudaFromNumpy(color_image)    
+    detections=net.Detect(cuda_image, color_image.shape[1], color_image.shape[0])
+    
+    print("#######################################")
+    print("Temps", time.time()-start)
+    for detection in detections:
+        print(detection)
+        x, y=detection.Center
+        x1=int(x-detection.Width/2)
+        y1=int(y-detection.Height/2)
+        x2=int(x+detection.Width/2)
+        y2=int(y+detection.Height/2)
+        if detection.ClassID==1:
+            cv2.rectangle(color_image, (x1, y1), (x2, y2), (255, 0, 0), 2)
+            dist=depth_frame.get_distance(int(x), int(y))
+            if dist<1:
+                msg="{:2.0f} cm".format(dist*100)
+            else:
+                msg="{:4.2f} m".format(dist)
+            cv2.putText(color_image, msg, (int(x), int(y)), cv2.FONT_HERSHEY_DUPLEX, 1, (255, 255, 255), 1, cv2.LINE_AA)
+            
+    cv2.imshow('RealSense1', depth_colormap)
+    cv2.imshow('RealSense2', color_image)
+
+    display.Render(cuda_image)
+    cuda_image=jetson.utils.cudaToNumpy(cuda_image)
+    #display.SetStatus("Object Detection | Network {:.0f} FPS".format(net.GetNetworkFPS()))
+    cv2.imshow('cuda_image', cuda_image)
+    key=cv2.waitKey(1)&0xFF
+    if key==ord('q'):
+        pipeline.stop()
+        quit()

Divers/odrive/README.md

@@ -0,0 +1,5 @@
+# Parlons actuator, odrive, moteur ...
+## Exemple de code avec la carte odrive robotics
+
+La vidéo de ce tutoriel est disponible à l'adresse suivante: https://www.youtube.com/watch?v=Mg-KiG3Rq2Q
+

Divers/odrive/test0.py

@@ -0,0 +1,49 @@
+import time
+import odrive
+from odrive.enums import *
+
+accel=30.
+vel=4.
+calibration=False
+
+odrv0=odrive.find_any(serial_number='205C3690424D')
+
+if calibration:
+    print("Calibration...", end='', flush=True)
+    odrv0.axis0.requested_state=4
+    odrv0.axis1.requested_state=4
+
+    while odrv0.axis0.current_state != AXIS_STATE_IDLE:
+        time.sleep(0.1)
+    while odrv0.axis1.current_state != AXIS_STATE_IDLE:
+        time.sleep(0.1)        
+    print("OK")
+    
+odrv0.axis0.controller.config.input_mode=INPUT_MODE_TRAP_TRAJ
+odrv0.axis0.trap_traj.config.vel_limit=vel
+odrv0.axis0.trap_traj.config.accel_limit=accel
+odrv0.axis0.trap_traj.config.decel_limit=accel
+odrv0.axis0.controller.config.inertia=0
+
+odrv0.axis1.controller.config.input_mode=INPUT_MODE_TRAP_TRAJ
+odrv0.axis1.trap_traj.config.vel_limit=vel
+odrv0.axis1.trap_traj.config.accel_limit=accel
+odrv0.axis1.trap_traj.config.decel_limit=accel
+odrv0.axis1.controller.config.inertia=0
+
+odrv0.axis0.requested_state=AXIS_STATE_CLOSED_LOOP_CONTROL
+odrv0.axis1.requested_state=AXIS_STATE_CLOSED_LOOP_CONTROL
+
+pos0=odrv0.axis0.encoder.pos_estimate
+pos1=odrv0.axis1.encoder.pos_estimate
+shift=1/5
+
+while True:
+    pos0_finale=pos0+shift
+    pos1_finale=pos1+shift
+    odrv0.axis0.controller.input_pos=pos0_finale
+    odrv0.axis1.controller.input_pos=pos1_finale
+    while abs(odrv0.axis0.encoder.pos_estimate-pos0_finale)>0.02 or \
+          abs(odrv0.axis1.encoder.pos_estimate-pos1_finale)>0.02:
+        time.sleep(0.1)
+    shift=-shift

Divers/odrive/test1.py

@@ -0,0 +1,55 @@
+import time
+import odrive
+from odrive.enums import *
+
+accel=10.
+vel=3.
+calibration=False
+
+odrv0=odrive.find_any(serial_number='205C3690424D')
+
+if calibration:
+    print("Calibration...", end='', flush=True)
+    odrv0.axis0.requested_state=4
+    odrv0.axis1.requested_state=4
+
+    while odrv0.axis0.current_state != AXIS_STATE_IDLE:
+        time.sleep(0.1)
+    while odrv0.axis1.current_state != AXIS_STATE_IDLE:
+        time.sleep(0.1)
+        
+    print("OK")
+
+odrv0.axis0.controller.config.control_mode = CONTROL_MODE_VELOCITY_CONTROL
+odrv0.axis0.controller.config.input_mode = INPUT_MODE_VEL_RAMP
+odrv0.axis1.controller.config.control_mode = CONTROL_MODE_VELOCITY_CONTROL
+odrv0.axis1.controller.config.input_mode = INPUT_MODE_VEL_RAMP
+    
+odrv0.axis0.controller.config.vel_ramp_rate=accel
+odrv0.axis1.controller.config.vel_ramp_rate=accel
+
+odrv0.axis0.requested_state=AXIS_STATE_CLOSED_LOOP_CONTROL
+odrv0.axis1.requested_state=AXIS_STATE_CLOSED_LOOP_CONTROL
+
+odrv0.axis0.controller.input_vel=vel
+odrv0.axis1.controller.input_vel=0
+
+evenement=time.time()
+id=0
+while True:
+    if id==0:
+        torque=odrv0.axis0.motor.current_control.Iq_setpoint*odrv0.axis0.motor.config.torque_constant
+    else:
+        torque=odrv0.axis1.motor.current_control.Iq_setpoint*odrv0.axis1.motor.config.torque_constant
+    if abs(torque)>0.2 and (time.time()-evenement)>1:
+        if id==0:
+            odrv0.axis0.controller.input_vel=0
+            odrv0.axis1.controller.input_vel=vel
+            id=1
+        else:
+            odrv0.axis1.controller.input_vel=0
+            odrv0.axis0.controller.input_vel=vel
+            id=0
+        evenement0=time.time()
+            
+

Divers/odrive/test2.py

@@ -0,0 +1,40 @@
+import time
+import odrive
+from odrive.enums import *
+
+accel=20.
+vel=5.
+ratio=2
+calibration=False
+
+odrv0=odrive.find_any(serial_number='205C3690424D')
+
+if calibration:
+    print("Calibration...", end='', flush=True)
+    odrv0.axis0.requested_state=4
+    odrv0.axis1.requested_state=4
+
+    while odrv0.axis0.current_state != AXIS_STATE_IDLE:
+        time.sleep(0.1)
+    while odrv0.axis1.current_state != AXIS_STATE_IDLE:
+        time.sleep(0.1)
+        
+    print("OK")
+
+odrv0.axis0.requested_state=AXIS_STATE_CLOSED_LOOP_CONTROL
+odrv0.axis1.requested_state=AXIS_STATE_CLOSED_LOOP_CONTROL
+
+odrv0.axis0.controller.config.input_mode = INPUT_MODE_TRAP_TRAJ
+odrv0.axis0.trap_traj.config.vel_limit=vel
+odrv0.axis0.trap_traj.config.accel_limit=accel
+odrv0.axis0.trap_traj.config.decel_limit=accel
+odrv0.axis0.controller.config.inertia=0
+
+pos0=odrv0.axis0.encoder.pos_estimate
+pos1=odrv0.axis1.encoder.pos_estimate
+
+odrv0.axis1.requested_state=AXIS_STATE_IDLE
+
+while True:
+    delta1=odrv0.axis1.encoder.pos_estimate-pos1
+    odrv0.axis0.controller.input_pos=pos0+ratio*delta1

Divers/renforcement1/README.md

@@ -0,0 +1,5 @@
+# Apprentissage par renforcement
+## Processus de décision markovien
+
+La vidéo de ce tutoriel est disponible à l'adresse suivante:<br>
+https://www.youtube.com/watch?v=Rgxs8lfoG4I

Divers/renforcement1/q_valeur.py

@@ -0,0 +1,35 @@
+import numpy as np
+import time
+
+Q=[[0, 0], [0, 0], [0, 0]]
+
+T=[[[0.50, 0.00, 0.50], [0.00, 0.00, 1.00]],
+   [[0.70, 0.10, 0.20], [0.00, 0.95, 0.05]],
+   [[0.40, 0.00, 0.60], [0.30, 0.30, 0.40]]]
+
+R=[[[ 0.00,  0.00, 0.00], [ 0.00,  0.00,  0.00]],
+   [[+5.00,  0.00, 0.00], [ 0.00,  0.00,  0.00]],
+   [[ 0.00,  0.00, 0.00], [-1.00,  0.00,  0.00]]]
+
+gamma=0.95
+
+for i in range(200):
+    time.sleep(0.05)
+    tab_somme_action=[]
+    for S in range(3):
+        for A in range(2):
+            somme=0
+            for s in range(3):
+                somme+=T[S][A][s]*(R[S][A][s]+gamma*np.max(Q[s]))
+            Q[S][A]=somme
+
+    print("---------------------------------")
+    print("Iteration:", i)
+    for S in range(3):
+        print()
+        for A in range(2):
+            text="Q[etat:{}, action:{}]={:+10.4f}".format(S, A, Q[S][A])
+            if A==np.argmax(Q[S]):
+                text=text+" <-"
+            print(text)
+print("---------------------------------")

Divers/renforcement2/CartPole_common.py

@@ -0,0 +1,12 @@
+import numpy as np
+
+# Valeurs hautes et basses des observations
+low_values=np.array([-5, -5, -0.45, -5])
+high_values=np.array([5, 5, 0.45, 5])
+
+division=[42, 42, 42, 42]
+pas=(high_values-low_values)/division
+
+def discretise(state):
+    discrete_state=(state-low_values)/pas
+    return tuple(discrete_state.astype(np.int))

Divers/renforcement2/CartPole_predict.py

@@ -0,0 +1,25 @@
+import gym
+import numpy as np
+import CartPole_common
+
+env=gym.make("CartPole-v0")
+env._max_episode_steps=5000
+
+q_table=np.load("CartPole_qtable.npy")
+
+for epoch in range(1000):
+    state = env.reset()
+    score = 0
+    while True:
+        env.render()
+        discrete_state=CartPole_common.discretise(state)
+        action=np.argmax(q_table[discrete_state])
+        #if not np.random.randint(5):
+        #    action=np.random.randint(2)
+        state, reward, done, info=env.step(action)
+        score+=reward
+        if done:
+            print('Essai {:05d} Score: {:04d}'.format(epoch, int(score)))
+            break
+
+env.close()

Divers/renforcement2/CartPole_train.py

@@ -0,0 +1,82 @@
+import gym
+import numpy as np
+import cv2
+import CartPole_common
+
+env=gym.make("CartPole-v0")
+env._max_episode_steps=500
+
+alpha=0.05
+gamma=0.98
+
+epoch=50000
+show_every=500
+
+epsilon=1.
+epsilon_min=0.05
+start_epsilon=1
+end_epsilon=epoch//2
+epsilon_decay_value=epsilon/(end_epsilon-start_epsilon)
+
+nbr_action=env.action_space.n
+q_table=np.random.uniform(low=-1, high=1, size=(CartPole_common.division+[nbr_action]))
+
+result_done=0
+scores=[]
+best_score=0
+for episode in range(epoch):
+    obs=env.reset()
+    discrete_state=CartPole_common.discretise(obs)
+    done=False
+
+    if episode%show_every == 0:
+        render=True
+        mean_score=np.mean(scores)
+        print("Epoch {:06d}/{:06d} reussite:{:04d}/{:04d} epsilon={:06.4f} Mean score={:08.4f} alpha={:06.4f}".format(episode, epoch, result_done, show_every, epsilon, mean_score, alpha))
+        scores=[]
+        result_done=0
+        if mean_score>best_score:
+            print("Sauvegarde ...")
+            np.save("CartPole_qtable", q_table)
+            best_score=mean_score
+        alpha=alpha*0.99
+
+    else:
+        render=False
+
+    score=1
+    while not done:
+
+        if np.random.random()>epsilon:
+            action=np.argmax(q_table[discrete_state])
+        else:
+            action=np.random.randint(nbr_action)
+
+        new_state, reward, done, info=env.step(action)
+        new_discrete_state=CartPole_common.discretise(new_state)
+
+        if episode%show_every == 0:
+            env.render()
+
+        #reward=2-np.abs(new_state[0])
+        if done:
+            scores.append(score)
+            if score==env._max_episode_steps:
+                result_done+=1
+            else:
+                reward=-10
+
+        max_future_q=np.max(q_table[new_discrete_state])
+        current_q=q_table[discrete_state][action]
+        new_q=(1-alpha)*current_q+alpha*(reward+gamma*max_future_q)
+        q_table[discrete_state][action]=new_q
+
+        score+=1
+        discrete_state=new_discrete_state
+
+    if end_epsilon>=episode>=start_epsilon:
+        epsilon-=epsilon_decay_value
+        if epsilon<epsilon_min:
+            epsilon=epsilon_min
+
+env.close()

Divers/renforcement2/MountainCar_common.py

@@ -0,0 +1,12 @@
+import numpy as np
+
+# Valeurs hautes et basses des observations
+low_values =np.array([-1.2, -0.07])
+high_values=np.array([0.6, 0.07])
+
+division=[42, 42]
+pas=(high_values-low_values)/division
+
+def discretise(state):
+    discrete_state=(state-low_values)/pas
+    return tuple(discrete_state.astype(np.int))

Divers/renforcement2/MountainCar_predict.py

@@ -0,0 +1,19 @@
+import gym
+import numpy as np
+import MountainCar_common
+
+env=gym.make("MountainCar-v0")
+
+q_table=np.load("MountainCar_qtable.npy")
+
+for epoch in range(1000):
+    state = env.reset()
+    while True:
+        env.render()
+        discrete_state=MountainCar_common.discretise(state)
+        action=np.argmax(q_table[discrete_state])
+        state, reward, done, info=env.step(action)
+        if done:
+            print("Essai {:05d}: {}".format(epoch, "OK" if state[0]>=env.goal_position else "raté ..."))
+            break
+env.close()

Divers/renforcement2/MountainCar_train.py

@@ -0,0 +1,68 @@
+import gym
+import numpy as np
+import MountainCar_common
+
+env=gym.make("MountainCar-v0")
+
+# Coefficient d'apprentissage
+alpha=0.1
+# Le "discount rate"
+gamma=0.98
+
+epoch=25000
+show_every=500
+
+# Politique exploration/exploitation
+epsilon=1.
+epsilon_min=0.1
+start_epsilon=1
+end_epsilon=epoch//2
+epsilon_decay_value=epsilon/(end_epsilon-start_epsilon)
+
+nbr_action=env.action_space.n
+q_table=np.random.uniform(low=-1, high=1, size=(MountainCar_common.division+[nbr_action]))
+
+OK=0
+for episode in range(epoch):
+    obs=env.reset()
+    discrete_state=MountainCar_common.discretise(obs)
+    done=False
+
+    if episode%show_every == 0:
+        render=True
+        print("epoch {:06d}/{:06d}  reussite:{:04d}/{:04d}  epsilon={:08.6f}".format(episode, epoch, OK, show_every, epsilon))
+        OK=0
+    else:
+        render=False
+
+    while not done:
+
+        if np.random.random()>epsilon:
+            action=np.argmax(q_table[discrete_state])
+        else:
+            action=np.random.randint(nbr_action)
+
+        new_state, reward, done, info=env.step(action)
+        new_discrete_state=MountainCar_common.discretise(new_state)
+        if episode%show_every == 0:
+            env.render()
+
+        if new_state[0]>=env.goal_position:
+            reward=1
+            OK+=1
+
+        # Mise à jour de Q(s, a) avec la formule de Bellman
+        max_future_q=np.max(q_table[new_discrete_state])
+        current_q=q_table[discrete_state][action]
+        new_q=(1-alpha)*current_q+alpha*(reward+gamma*max_future_q)
+        q_table[discrete_state][action]=new_q
+
+        discrete_state=new_discrete_state
+
+    if end_epsilon>=episode>=start_epsilon:
+        epsilon-=epsilon_decay_value
+        if epsilon<epsilon_min:
+            epsilon=epsilon_min
+
+np.save("MountainCar_qtable", q_table)
+env.close()

Divers/renforcement2/README.md

@@ -0,0 +1,6 @@
+# Apprentissage par renforcement
+## Equation de Bellman
+
+La vidéo de ce tutoriel est disponible à l'adresse suivante:<br>
+https://www.youtube.com/watch?v=4Ak6OyehqJc
+

Divers/renforcement3/README.md

@@ -0,0 +1,7 @@
+# Apprentissage par renforcement
+## Q learning "basique" avec un perceptron
+
+La vidéo de ce tutoriel est disponible à l'adresse suivante:<br>
+https://www.youtube.com/watch?v=03U-3BOqfMs
+
+

Divers/renforcement3/graph.py

@@ -0,0 +1,25 @@
+import numpy as np
+import matplotlib.pyplot as plot
+import sys
+
+fenetre=50
+max_score=500
+
+if len(sys.argv)!=2:
+    print("Usage:", sys.argv[0], "<fichier npy>")
+    quit()
+
+tab_s=np.load(sys.argv[1])
+
+tab_m=[]
+for i in range(len(tab_s)-fenetre):
+    m=np.mean(tab_s[i:i+fenetre])
+    tab_m.append(m)
+
+fig=plot.gcf()
+fig.set_size_inches(12, 6)
+plot.plot(tab_s)
+plot.grid()
+plot.ylim(0, max_score)
+plot.plot(np.arange(fenetre, len(tab_s)), tab_m, color='#FF0000')
+plot.show()

Divers/renforcement3/predict.py

@@ -0,0 +1,21 @@
+import gym
+import tensorflow as tf
+import numpy as np
+
+env=gym.make("CartPole-v0")
+env._max_episode_steps=500
+
+model=tf.keras.models.load_model("my_model")
+
+while True:
+  observations=env.reset()
+  score=0
+  while True:
+    env.render()
+    valeurs_q=model(np.expand_dims(observations, axis=0))
+    action=int(tf.argmax(valeurs_q[0], axis=-1))
+    observations, reward, done, info=env.step(action)
+    if done:
+      print("SCORE", score)
+      break
+    score+=1

Divers/renforcement3/train.py

@@ -0,0 +1,108 @@
+import gym
+import tensorflow as tf
+from tensorflow.keras import models, layers
+import numpy as np
+
+env = gym.make("CartPole-v0")
+env._max_episode_steps=500
+nbr_action=2
+
+gamma=tf.constant(0.98)
+epoch=20000
+best_score=0
+
+epsilon=1.
+epsilon_min=0.10
+start_epsilon=1
+end_epsilon=epoch//2
+epsilon_decay_value=epsilon/(end_epsilon-start_epsilon)
+
+def model():
+  entree=layers.Input(shape=(4), dtype='float32')
+  result=layers.Dense(30, activation='relu')(entree)
+  result=layers.Dense(30, activation='relu')(result)
+  sortie=layers.Dense(nbr_action)(result)
+    
+  model=models.Model(inputs=entree, outputs=sortie)
+  return model
+
+def my_loss(target_q, predicted_q):
+  loss=tf.reduce_mean(tf.math.square(target_q-predicted_q))
+  return loss
+
+@tf.function
+def train_step(reward, action, observation, next_observation, done):
+  next_Q_values=model(next_observation)
+  best_next_actions=tf.math.argmax(next_Q_values, axis=1)
+  next_mask=tf.one_hot(best_next_actions, nbr_action)
+  next_best_Q_values=tf.reduce_sum(next_Q_values*next_mask, axis=1)
+  target_Q_values=reward+(1-done)*gamma*next_best_Q_values
+  target_Q_values=tf.reshape(target_Q_values, (-1, 1))
+  mask=tf.one_hot(action, nbr_action)
+  with tf.GradientTape() as tape:
+    all_Q_values=model(observation)
+    Q_values=tf.reduce_sum(all_Q_values*mask, axis=1, keepdims=True)
+    loss=my_loss(target_Q_values, Q_values)
+  gradients=tape.gradient(loss, model.trainable_variables)
+  optimizer.apply_gradients(zip(gradients, model.trainable_variables))
+  train_loss(loss)
+
+def train(debug=False):
+  global epsilon, best_score, tab_score
+  for e in range(epoch):
+    print("EPOCH:", e, "epsilon", epsilon)
+    score=0
+    tab_observations=[]
+    tab_rewards=[]
+    tab_actions=[]
+    tab_next_observations=[]
+    tab_done=[]
+    
+    observations=env.reset()
+    while True:
+      tab_observations.append(observations)    
+      if np.random.random()>epsilon:
+        valeurs_q=model(np.expand_dims(observations, axis=0))
+        action=int(tf.argmax(valeurs_q[0], axis=-1))
+      else:
+        action=np.random.randint(0, nbr_action)
+      observations, reward, done, info=env.step(action)
+      tab_actions.append(action)
+      tab_next_observations.append(observations)
+      tab_done.append(done)
+      if done:
+        tab_rewards.append(-10.)
+        print("FIN, score:", score)
+        tab_score.append(score)
+        score=0
+        break
+      score+=1
+      tab_rewards.append(reward)
+
+    tab_rewards=np.array(tab_rewards, dtype=np.float32)
+    tab_actions=np.array(tab_actions, dtype=np.int32)
+    tab_observations=np.array(tab_observations, dtype=np.float32)
+    tab_next_observations=np.array(tab_next_observations, dtype=np.float32)
+    tab_done=np.array(tab_done, dtype=np.float32)
+    train_step(tab_rewards, tab_actions, tab_observations, tab_next_observations, tab_done)
+    train_loss.reset_states()
+
+    epsilon-=epsilon_decay_value
+    epsilon=max(epsilon, epsilon_min)
+    if np.mean(tab_score[-20:])>best_score:
+      print("Sauvegarde du modele")
+      model.save("my_model")
+      best_score=np.mean(tab_score[-20:])
+      if best_score==env._max_episode_steps-1:
+        return
+      
+model=model()
+optimizer=tf.keras.optimizers.Adam(learning_rate=1E-4)
+train_loss=tf.keras.metrics.Mean()
+tab_s=[]
+
+tab_score=[]
+train()
+
+np.save("tab_score", tab_score)
+

Divers/renforcement3/train_better.py

@@ -0,0 +1,108 @@
+import gym
+import tensorflow as tf
+from tensorflow.keras import models, layers
+import numpy as np
+
+env=gym.make("CartPole-v0")
+env._max_episode_steps=200
+nbr_action=2
+
+fichier_log=open("log_critic.csv", "a")
+
+gamma=0.98
+max_episode=600
+epsilon=1.
+epsilon_min=0.10
+start_epsilon=10
+end_epsilon=max_episode
+epsilon_decay_value=epsilon/(end_epsilon-start_epsilon)
+
+def model():
+  entree=layers.Input(shape=(4), dtype='float32')
+  result=layers.Dense(32, activation='relu')(entree)
+  result=layers.Dense(32, activation='relu')(result)
+  sortie=layers.Dense(nbr_action)(result)
+
+  model=models.Model(inputs=entree, outputs=sortie)
+  return model
+
+def my_loss(target_q, predicted_q):
+  loss=tf.reduce_mean(tf.math.square(target_q-predicted_q))
+  return loss
+
+@tf.function
+def train_step(reward, action, observation, next_observation, done):
+  next_Q_values=model(next_observation)
+  best_next_actions=tf.math.argmax(next_Q_values, axis=1)
+  next_mask=tf.one_hot(best_next_actions, nbr_action)
+  next_best_Q_values=tf.reduce_sum(next_Q_values*next_mask, axis=1)
+  target_Q_values=reward+(1-done)*gamma*next_best_Q_values
+  target_Q_values=tf.reshape(target_Q_values, (-1, 1))
+  mask=tf.one_hot(action, nbr_action)
+  with tf.GradientTape() as tape:
+    all_Q_values=model(observation)
+    Q_values=tf.reduce_sum(all_Q_values*mask, axis=1, keepdims=True)
+    loss=my_loss(target_Q_values, Q_values)
+  gradients=tape.gradient(loss, model.trainable_variables)
+  optimizer.apply_gradients(zip(gradients, model.trainable_variables))
+  train_loss(loss)
+
+def train(debug=False):
+  global epsilon
+  m_reward=0
+  for episode in range(max_episode):
+    score=0
+    tab_observations=[]
+    tab_rewards=[]
+    tab_actions=[]
+    tab_next_observations=[]
+    tab_done=[]
+
+    observations=env.reset()
+    score=0
+    while True:
+      tab_observations.append(observations)
+      if np.random.random()>epsilon:
+        valeurs_q=model(np.expand_dims(observations, axis=0))
+        action=int(tf.argmax(valeurs_q[0], axis=-1))
+      else:
+        action=np.random.randint(0, nbr_action)
+      observations, reward, done, info=env.step(action)
+      score+=reward
+      tab_actions.append(action)
+      tab_next_observations.append(observations)
+      tab_done.append(done)
+      if done:
+        tab_rewards.append(-10.)
+        break
+      tab_rewards.append(reward)
+
+    tab_rewards=np.array(tab_rewards, dtype=np.float32)
+    tab_actions=np.array(tab_actions, dtype=np.int32)
+    tab_observations=np.array(tab_observations, dtype=np.float32)
+    tab_next_observations=np.array(tab_next_observations, dtype=np.float32)
+    tab_done=np.array(tab_done, dtype=np.float32)
+    train_step(tab_rewards, tab_actions, tab_observations, tab_next_observations, tab_done)
+    train_loss.reset_states()
+
+    epsilon-=epsilon_decay_value
+    epsilon=max(epsilon, epsilon_min)
+
+    m_reward=0.05*score+(1-0.05)*m_reward
+    message="Episode {:04d}  score:{:6.1f}  moyenne lissée: {:6.1f} (epsilon={:5.3f})"
+    print(message.format(episode, score, m_reward, epsilon))
+
+    fichier_log.write("{:f}:{:f}\n".format(score, m_reward))
+
+    if m_reward>env._max_episode_steps-10:
+        print("Fin de l'apprentissage".format(episode))
+        break
+
+model=model()
+optimizer=tf.keras.optimizers.Adam(learning_rate=1E-2)
+train_loss=tf.keras.metrics.Mean()
+tab_s=[]
+
+train()
+
+fichier_log.close()

Divers/renforcement4/README.md

@@ -0,0 +1,7 @@
+# Apprentissage par renforcement
+## Pacman en mode target !
+
+La vidéo de ce tutoriel est disponible à l'adresse suivante:<br>
+https://www.youtube.com/watch?v=F-u9AOMt7zo
+
+![alt text](https://github.com/L42Project/Tutoriels/blob/master/Divers/renforcement4/img.png)

Divers/renforcement4/joue.py

@@ -0,0 +1,72 @@
+import gym
+import cv2
+import tensorflow as tf
+from tensorflow.keras import models, layers
+import numpy as np
+import time
+import matplotlib.pyplot as plot
+import os
+
+env=gym.make("MsPacman-v0")
+
+#model=tf.keras.models.load_model('my_model_v1')
+model=tf.keras.models.load_model('my_model_target')
+
+decalage_debut=90
+taille_sequence=6
+
+def transform_img(image):
+  result=np.expand_dims(image[:170, :, 0], axis=-1)
+  return result
+
+def joue():
+
+  ######
+  observations=env.reset()
+  vie=3
+  for i in range(decalage_debut-taille_sequence):
+    env.step(0)
+  tab_sequence=[]
+  for i in range(taille_sequence):
+    observation, reward, done, info=env.step(0)
+    img=transform_img(observation)
+    tab_sequence.append(img)
+  tab_sequence=np.array(tab_sequence, dtype=np.float32)
+  ######
+
+  tab_img=[]
+  score=0
+  vie=3
+  while True:
+    valeurs_q=model(np.expand_dims(np.concatenate(tab_sequence, axis=-1), axis=0))
+    action=int(tf.argmax(valeurs_q[0], axis=-1))
+    print(action+1, end=' ')
+    score+=min(reward, 10.)
+    if info['ale.lives']<vie:
+      print("XXX", end=" ")
+      vie-=1
+    if done:
+      print("\nSCORE:", score)
+      return tab_img, score
+    
+    observation, reward, done, info=env.step(action+1)
+    if reward>10:
+      print("MIAM", reward, end=" ")
+
+    img=transform_img(observation)
+    tab_sequence[:-1]=tab_sequence[1:]
+    tab_sequence[taille_sequence-1]=img
+    tab_img.append(observation)
+    
+score=0
+while score<1400:
+  start_time=time.time()
+  tab_img, score=joue()
+  print(time.time()-start_time)
+      
+for i in range(len(tab_img)):
+  cv2.imshow("Pacman", tab_img[i])
+  key=cv2.waitKey(20)
+  if key==ord('q'):
+    break
+    

Divers/renforcement4/train_target.py

@@ -0,0 +1,180 @@
+import gym
+import cv2
+import tensorflow as tf
+from tensorflow.keras import models, layers
+import numpy as np
+import time
+import matplotlib.pyplot as plot
+
+env = gym.make("MsPacman-v0")
+print("Liste des actions", env.unwrapped.get_action_meanings())
+nbr_action=tf.constant(4)
+
+file_model='my_model_target'
+file_stats='tab_score_target'
+
+gamma=tf.constant(0.999)
+epoch=200
+decalage_debut=90
+taille_sequence=6
+nbr_jeu=300
+pourcentage_batch=0.20
+best_score=0
+
+epsilon=1.
+epsilon_min=0.10
+start_epsilon=1
+end_epsilon=epoch//4
+epsilon_decay_value=epsilon/(end_epsilon-start_epsilon)
+
+def model(nbr_cc=8):
+  entree=layers.Input(shape=(170, 160, taille_sequence), dtype='float32')
+  result=layers.Conv2D(  nbr_cc, 3, activation='relu', padding='same', strides=2)((entree/128)-1)
+  result=layers.Conv2D(2*nbr_cc, 3, activation='relu', padding='same', strides=2)(result)
+  result=layers.BatchNormalization()(result)
+  result=layers.Conv2D(4*nbr_cc, 3, activation='relu', padding='same', strides=2)(result)
+  result=layers.Conv2D(8*nbr_cc, 3, activation='relu', padding='same', strides=2)(result)
+  result=layers.BatchNormalization()(result)
+
+  result=layers.Flatten()(result)
+
+  result=layers.Dense(512, activation='relu')(result)
+  sortie=layers.Dense(nbr_action)(result)
+    
+  model=models.Model(inputs=entree, outputs=sortie)
+  return model
+
+def transform_img(image):
+  result=np.expand_dims(image[:170, :, 0], axis=-1)
+  return result
+
+def simulation(epsilon, debug=False):
+  if debug:
+    start_time=time.time()
+
+  tab_observations=[]
+  tab_rewards=[]
+  tab_actions=[]
+  tab_next_observations=[]
+  tab_done=[]
+  
+  ######
+  observations=env.reset()
+  vie=3
+  for i in range(decalage_debut-taille_sequence):
+    env.step(0)
+  tab_sequence=[]
+  for i in range(taille_sequence):
+    observation, reward, done, info=env.step(0)
+    img=transform_img(observation)
+    tab_sequence.append(img)
+  tab_sequence=np.array(tab_sequence, dtype=np.float32)
+  ######
+
+  score=0
+  while True:
+    if np.random.random()>epsilon:
+      valeurs_q=model_primaire(np.expand_dims(np.concatenate(tab_sequence, axis=-1), axis=0))
+      action=int(tf.argmax(valeurs_q[0], axis=-1))
+    else:
+      action=np.random.randint(0, nbr_action)
+
+    h=np.random.randint(10)
+    if h==0:
+      tab_observations.append(np.concatenate(tab_sequence, axis=-1))
+      tab_actions.append(action)
+    score+=reward
+    if info['ale.lives']<vie:
+      reward=-50.
+      vie=info['ale.lives']
+      if h==0:
+        tab_done.append(True)
+    else:
+      if h==0:
+        tab_done.append(done)
+    if h==0:
+      tab_rewards.append(reward)
+    if done:
+      tab_s.append(score)
+      if h==0:
+        tab_sequence[:-1]=tab_sequence[1:]
+        tab_sequence[taille_sequence-1]=img
+        tab_next_observations.append(np.concatenate(tab_sequence, axis=-1))
+      tab_done=np.array(tab_done, dtype=np.float32)
+      tab_observations=np.array(tab_observations, dtype=np.float32)
+      tab_next_observations=np.array(tab_next_observations, dtype=np.float32)
+      tab_rewards=np.array(tab_rewards, dtype=np.float32)
+      tab_rewards[tab_rewards==0]=-1.
+      tab_rewards[tab_rewards>10]=10.
+      tab_actions=np.array(tab_actions, dtype=np.int32)
+      if debug:
+            print("  Creation observations {:5.3f} seconde(s)".format(float(time.time()-start_time)))
+            print("     score:{:5d}   batch:{:4d}".format(int(score), len(tab_done)))
+      return tab_observations,\
+             tab_rewards,\
+             tab_actions,\
+             tab_next_observations,\
+             tab_done
+    observation, reward, done, info=env.step(action+1)
+    img=transform_img(observation)
+    tab_sequence[:-1]=tab_sequence[1:]
+    tab_sequence[taille_sequence-1]=img
+    if h==0:
+      tab_next_observations.append(np.concatenate(tab_sequence, axis=-1))
+
+def my_loss(y, q):
+  loss=tf.reduce_mean(tf.math.square(y-q))
+  return loss
+
+@tf.function
+def train_step(reward, action, observation, next_observation, done):
+  next_Q_values=model_cible(next_observation)
+  best_next_actions=tf.math.argmax(next_Q_values, axis=1)
+  next_mask=tf.one_hot(best_next_actions, nbr_action)
+  next_best_Q_values=tf.reduce_sum(next_Q_values*next_mask, axis=1)
+  target_Q_values=reward+(1-done)*gamma*next_best_Q_values
+  target_Q_values=tf.reshape(target_Q_values, (-1, 1))
+  mask=tf.one_hot(action, nbr_action)
+  with tf.GradientTape() as tape:
+    all_Q_values=model_primaire(observation)
+    Q_values=tf.reduce_sum(all_Q_values*mask, axis=1, keepdims=True)
+    loss=my_loss(target_Q_values, Q_values)
+  gradients=tape.gradient(loss, model_primaire.trainable_variables)
+  optimizer.apply_gradients(zip(gradients, model_primaire.trainable_variables))
+  train_loss(loss)
+  
+def train(debug=False):
+  global epsilon, best_score
+  for e in range(epoch):
+    for i in range(nbr_jeu):
+      print("Epoch {:04d}/{:05d} epsilon={:05.3f}".format(i, e, epsilon))
+      tab_observations, tab_rewards, tab_actions, tab_next_observations, tab_done=simulation(epsilon, debug=True)
+      if debug:
+        start_time=time.time()
+      train_step(tab_rewards, tab_actions, tab_observations, tab_next_observations, tab_done)
+      if debug:
+        print("  Entrainement {:5.3f} seconde(s)".format(float(time.time()-start_time)))
+        print("     loss: {:6.4f}".format(train_loss.result()))
+      train_loss.reset_states()
+
+    print("Copie des poids primaire -> cible")
+    for a, b in zip(model_cible.variables, model_primaire.variables):
+      a.assign(b)
+    
+    epsilon-=epsilon_decay_value
+    epsilon=max(epsilon, epsilon_min)
+    np.save(file_stats, tab_s)
+    if np.mean(tab_s[-200:])>best_score:
+      print("Sauvegarde du modele")
+      model_cible.save(file_model)
+      best_score=np.mean(tab_s[-200:])
+
+model_primaire=model(16)
+model_cible=tf.keras.models.clone_model(model_primaire)
+for a, b in zip(model_cible.variables, model_primaire.variables):
+  a.assign(b)
+
+optimizer=tf.keras.optimizers.Adam(learning_rate=1E-4)
+train_loss=tf.keras.metrics.Mean()
+tab_s=[]
+train(debug=True)

Divers/renforcement4/train_v1.py

@@ -0,0 +1,173 @@
+import gym
+import cv2
+import tensorflow as tf
+from tensorflow.keras import models, layers
+import numpy as np
+import time
+import matplotlib.pyplot as plot
+
+env = gym.make("MsPacman-v0")
+print("Liste des actions", env.unwrapped.get_action_meanings())
+nbr_action=tf.constant(4)
+
+file_model='my_model_v1'
+file_stats='tab_score_v1'
+
+gamma=tf.constant(0.999)
+epoch=1500
+decalage_debut=90
+taille_sequence=6
+nbr_jeu=40
+pourcentage_batch=0.20
+best_score=0
+
+epsilon=1.
+epsilon_min=0.10
+start_epsilon=1
+end_epsilon=epoch//4
+epsilon_decay_value=epsilon/(end_epsilon-start_epsilon)
+
+def model(nbr_cc=8):
+  entree=layers.Input(shape=(170, 160, taille_sequence), dtype='float32')
+  result=layers.Conv2D(  nbr_cc, 3, activation='relu', padding='same', strides=2)((entree/128)-1)
+  result=layers.Conv2D(2*nbr_cc, 3, activation='relu', padding='same', strides=2)(result)
+  result=layers.BatchNormalization()(result)
+  result=layers.Conv2D(4*nbr_cc, 3, activation='relu', padding='same', strides=2)(result)
+  result=layers.Conv2D(8*nbr_cc, 3, activation='relu', padding='same', strides=2)(result)
+  result=layers.BatchNormalization()(result)
+
+  result=layers.Flatten()(result)
+
+  result=layers.Dense(512, activation='relu')(result)
+  sortie=layers.Dense(nbr_action)(result)
+    
+  model=models.Model(inputs=entree, outputs=sortie)
+  return model
+
+def transform_img(image):
+  result=np.expand_dims(image[:170, :, 0], axis=-1)
+  return result
+
+def simulation(epsilon, debug=False):
+  if debug:
+    start_time=time.time()
+
+  tab_observations=[]
+  tab_rewards=[]
+  tab_actions=[]
+  tab_next_observations=[]
+  tab_done=[]
+  
+  ######
+  observations=env.reset()
+  vie=3
+  for i in range(decalage_debut-taille_sequence):
+    env.step(0)
+  tab_sequence=[]
+  for i in range(taille_sequence):
+    observation, reward, done, info=env.step(0)
+    img=transform_img(observation)
+    tab_sequence.append(img)
+  tab_sequence=np.array(tab_sequence, dtype=np.float32)
+  ######
+
+  score=0
+  while True:
+    if np.random.random()>epsilon:
+      valeurs_q=model(np.expand_dims(np.concatenate(tab_sequence, axis=-1), axis=0))
+      action=int(tf.argmax(valeurs_q[0], axis=-1))
+    else:
+      action=np.random.randint(0, nbr_action)
+
+    h=np.random.randint(10)
+    if h==0:
+      tab_observations.append(np.concatenate(tab_sequence, axis=-1))
+      tab_actions.append(action)
+    score+=reward
+    if info['ale.lives']<vie:
+      reward=-50.
+      vie=info['ale.lives']
+      if h==0:
+        tab_done.append(True)
+    else:
+      if h==0:
+        tab_done.append(done)
+    if h==0:
+      tab_rewards.append(reward)
+    if done:
+      tab_s.append(score)
+      if h==0:
+        tab_sequence[:-1]=tab_sequence[1:]
+        tab_sequence[taille_sequence-1]=img
+        tab_next_observations.append(np.concatenate(tab_sequence, axis=-1))
+      tab_done=np.array(tab_done, dtype=np.float32)
+      tab_observations=np.array(tab_observations, dtype=np.float32)
+      tab_next_observations=np.array(tab_next_observations, dtype=np.float32)
+      tab_rewards=np.array(tab_rewards, dtype=np.float32)
+      tab_rewards[tab_rewards==0]=-1.
+      tab_rewards[tab_rewards>10]=10.
+      tab_actions=np.array(tab_actions, dtype=np.int32)
+      if debug:
+            print("  Creation observations {:5.3f} seconde(s)".format(float(time.time()-start_time)))
+            print("     score:{:5d}   batch:{:4d}".format(int(score), len(tab_done)))
+      return tab_observations,\
+             tab_rewards,\
+             tab_actions,\
+             tab_next_observations,\
+             tab_done
+    observation, reward, done, info=env.step(action+1)
+    img=transform_img(observation)
+    tab_sequence[:-1]=tab_sequence[1:]
+    tab_sequence[taille_sequence-1]=img
+    if h==0:
+      tab_next_observations.append(np.concatenate(tab_sequence, axis=-1))
+
+def my_loss(y, q):
+  loss=tf.reduce_mean(tf.math.square(y-q))
+  return loss
+
+@tf.function
+def train_step(reward, action, observation, next_observation, done):
+  next_Q_values=model(next_observation)
+  best_next_actions=tf.math.argmax(next_Q_values, axis=1)
+  next_mask=tf.one_hot(best_next_actions, nbr_action)
+  next_best_Q_values=tf.reduce_sum(next_Q_values*next_mask, axis=1)
+  target_Q_values=reward+(1-done)*gamma*next_best_Q_values
+  target_Q_values=tf.reshape(target_Q_values, (-1, 1))
+  mask=tf.one_hot(action, nbr_action)
+  with tf.GradientTape() as tape:
+    all_Q_values=model(observation)
+    Q_values=tf.reduce_sum(all_Q_values*mask, axis=1, keepdims=True)
+    loss=my_loss(target_Q_values, Q_values)
+  gradients=tape.gradient(loss, model.trainable_variables)
+  optimizer.apply_gradients(zip(gradients, model.trainable_variables))
+  train_loss(loss)
+  
+def train(debug=False):
+  global epsilon, best_score
+  for e in range(epoch):
+    for i in range(nbr_jeu):
+      print("Epoch {:04d}/{:05d} epsilon={:05.3f}".format(i, e, epsilon))
+      tab_observations, tab_rewards, tab_actions, tab_next_observations, tab_done=simulation(epsilon, debug=True)
+      if debug:
+        start_time=time.time()
+      train_step(tab_rewards, tab_actions, tab_observations, tab_next_observations, tab_done)
+      if debug:
+        print("  Entrainement {:5.3f} seconde(s)".format(float(time.time()-start_time)))
+        print("     loss: {:6.4f}".format(train_loss.result()))
+      train_loss.reset_states()
+
+    epsilon-=epsilon_decay_value
+    epsilon=max(epsilon, epsilon_min)
+    np.save(file_stats, tab_s)
+    if np.mean(tab_s[-200:])>best_score:
+      print("Sauvegarde du modele")
+      model.save(file_model)
+      best_score=np.mean(tab_s[-200:])
+
+model=model(16)
+
+optimizer=tf.keras.optimizers.Adam(learning_rate=1E-4)
+train_loss=tf.keras.metrics.Mean()
+tab_s=[]
+train(debug=True)

Divers/renforcement5/README.md

@@ -0,0 +1,11 @@
+# Apprentissage par renforcement
+## Méthode 'acteur'
+
+La vidéo de ce tutoriel est disponible à l'adresse suivante:<br>
+https://www.youtube.com/watch?v=LtRAgxRb5eQ
+
+Ci dessous, le graph de l'apprentissage sur l'environnement CartPole (https://gym.openai.com/envs/CartPole-v0/)<br>
+En bleu: Méthode 'critique'<br>
+En orange: Méthode 'acteur'<br>
+
+![image](mini.png)

Divers/renforcement5/cartpole_actor.py

@@ -0,0 +1,83 @@
+import gym
+import tensorflow as tf
+from tensorflow.keras import models, layers
+import numpy as np
+import os
+
+env=gym.make("CartPole-v0")
+env._max_episode_steps=200
+nbr_actions=2
+gamma=0.99
+max_episode=600
+
+prefix_log_file="log_actor"
+id_file=0
+while os.path.exists(prefix_log_file+str(id_file)+".csv"):
+    id_file+=1
+fichier_log=open(prefix_log_file+str(id_file)+".csv", "w")
+print("Création du fichier de log", prefix_log_file+str(id_file)+".csv")
+
+def model(nbr_inputs, nbr_hidden, nbr_actions):
+  entree=layers.Input(shape=(nbr_inputs), dtype='float32')
+  result=layers.Dense(32, activation='relu')(entree)
+  result=layers.Dense(32, activation='relu')(result)
+  sortie=layers.Dense(nbr_actions, activation='softmax')(result)
+
+  my_model=models.Model(inputs=entree, outputs=sortie)
+  return my_model
+
+def calcul_discount_rate(rewards_history, gamma, normalize=False):
+    result=[]
+    discounted_sum=0
+    for r in rewards_history[::-1]:
+        discounted_sum=r+gamma*discounted_sum
+        result.insert(0, discounted_sum)
+
+    # Normalisation
+    if normalize is True:
+        result=np.array(result)
+        result=(result-np.mean(result))/(np.std(result)+1E-7)
+        result=list(result)
+
+    return result
+
+def train():
+  m_reward=0
+  for episode in range(max_episode):
+    tab_rewards=[]
+    tab_prob_actions=[]
+
+    observations=env.reset()
+    with tf.GradientTape() as tape:
+        while True:
+                action_probs=my_model(np.expand_dims(observations, axis=0))
+                action=np.random.choice(nbr_actions, p=np.squeeze(action_probs))
+                tab_prob_actions.append(action_probs[0, action])
+                observations, reward, done, info=env.step(action)
+                tab_rewards.append(reward)
+                if done:
+                    break
+
+        discount_rate=calcul_discount_rate(tab_rewards, gamma, normalize=True)
+
+        loss=-tf.math.log(tab_prob_actions)*discount_rate
+        gradients=tape.gradient(loss, my_model.trainable_variables)
+        optimizer.apply_gradients(zip(gradients, my_model.trainable_variables))
+
+    score=sum(tab_rewards)
+    m_reward=0.05*score+(1-0.05)*m_reward
+    message="Episode {:04d}  score:{:6.1f}  MPE: {:6.1f}"
+    print(message.format(episode, score, m_reward))
+
+    fichier_log.write("{:f}:{:f}\n".format(score, m_reward))
+
+    if m_reward>env._max_episode_steps-10:
+        print("Fin de l'apprentissage".format(episode))
+        break
+
+my_model=model(4, 32, nbr_actions)
+optimizer=tf.keras.optimizers.Adam(learning_rate=1E-2)
+
+train()
+
+fichier_log.close()

Divers/renforcement5/cartpole_critic.py

@@ -0,0 +1,114 @@
+import gym
+import tensorflow as tf
+from tensorflow.keras import models, layers
+import numpy as np
+import os
+
+env=gym.make("CartPole-v0")
+env._max_episode_steps=200
+nbr_action=2
+
+prefix_log_file="log_critic_"
+id_file=0
+while os.path.exists(prefix_log_file+str(id_file)+".csv"):
+    id_file+=1
+fichier_log=open(prefix_log_file+str(id_file)+".csv", "w")
+print("Création du fichier de log", prefix_log_file+str(id_file)+".csv")
+
+gamma=0.98
+max_episode=600
+epsilon=1.
+epsilon_min=0.10
+start_epsilon=10
+end_epsilon=max_episode
+epsilon_decay_value=epsilon/(end_epsilon-start_epsilon)
+
+def model():
+  entree=layers.Input(shape=(4), dtype='float32')
+  result=layers.Dense(32, activation='relu')(entree)
+  result=layers.Dense(32, activation='relu')(result)
+  sortie=layers.Dense(nbr_action)(result)
+
+  model=models.Model(inputs=entree, outputs=sortie)
+  return model
+
+def my_loss(target_q, predicted_q):
+  loss=tf.reduce_mean(tf.math.square(target_q-predicted_q))
+  return loss
+
+@tf.function
+def train_step(reward, action, observation, next_observation, done):
+  next_Q_values=model(next_observation)
+  best_next_actions=tf.math.argmax(next_Q_values, axis=1)
+  next_mask=tf.one_hot(best_next_actions, nbr_action)
+  next_best_Q_values=tf.reduce_sum(next_Q_values*next_mask, axis=1)
+  target_Q_values=reward+(1-done)*gamma*next_best_Q_values
+  target_Q_values=tf.reshape(target_Q_values, (-1, 1))
+  mask=tf.one_hot(action, nbr_action)
+  with tf.GradientTape() as tape:
+    all_Q_values=model(observation)
+    Q_values=tf.reduce_sum(all_Q_values*mask, axis=1, keepdims=True)
+    loss=my_loss(target_Q_values, Q_values)
+  gradients=tape.gradient(loss, model.trainable_variables)
+  optimizer.apply_gradients(zip(gradients, model.trainable_variables))
+  train_loss(loss)
+
+def train(debug=False):
+  global epsilon
+  m_reward=0
+  for episode in range(max_episode):
+    score=0
+    tab_observations=[]
+    tab_rewards=[]
+    tab_actions=[]
+    tab_next_observations=[]
+    tab_done=[]
+
+    observations=env.reset()
+    score=0
+    while True:
+      tab_observations.append(observations)
+      if np.random.random()>epsilon:
+        valeurs_q=model(np.expand_dims(observations, axis=0))
+        action=int(tf.argmax(valeurs_q[0], axis=-1))
+      else:
+        action=np.random.randint(0, nbr_action)
+      observations, reward, done, info=env.step(action)
+      score+=reward
+      tab_actions.append(action)
+      tab_next_observations.append(observations)
+      tab_done.append(done)
+      if done:
+        tab_rewards.append(-10.)
+        break
+      tab_rewards.append(reward)
+
+    tab_rewards=np.array(tab_rewards, dtype=np.float32)
+    tab_actions=np.array(tab_actions, dtype=np.int32)
+    tab_observations=np.array(tab_observations, dtype=np.float32)
+    tab_next_observations=np.array(tab_next_observations, dtype=np.float32)
+    tab_done=np.array(tab_done, dtype=np.float32)
+    train_step(tab_rewards, tab_actions, tab_observations, tab_next_observations, tab_done)
+    train_loss.reset_states()
+
+    epsilon-=epsilon_decay_value
+    epsilon=max(epsilon, epsilon_min)
+
+    m_reward=0.05*score+(1-0.05)*m_reward
+    message="Episode {:04d}  score:{:6.1f}  MPE: {:6.1f} (epsilon={:5.3f})"
+    print(message.format(episode, score, m_reward, epsilon))
+
+    fichier_log.write("{:f}:{:f}\n".format(score, m_reward))
+
+    if m_reward>env._max_episode_steps-10:
+        print("Fin de l'apprentissage".format(episode))
+        break
+
+model=model()
+optimizer=tf.keras.optimizers.Adam(learning_rate=1E-2)
+train_loss=tf.keras.metrics.Mean()
+tab_s=[]
+
+train()
+
+fichier_log.close()

Divers/renforcement6/README.md

@@ -0,0 +1,12 @@
+# Apprentissage par renforcement
+## Méthode 'acteur/critique'
+
+La vidéo de ce tutoriel est disponible à l'adresse suivante:<br>
+https://www.youtube.com/watch?v=1okjkEMP79c
+
+Ci dessous, le graph de l'apprentissage sur l'environnement CartPole (https://gym.openai.com/envs/CartPole-v0/)<br>
+En bleu: Méthode 'critique'<br>
+En orange: Méthode 'acteur'<br>
+En vert: Méthode 'acteur/critique'<br>
+
+![image](graph.png)

Divers/renforcement6/cartpole_actor.py

@@ -0,0 +1,83 @@
+import gym
+import tensorflow as tf
+from tensorflow.keras import models, layers
+import numpy as np
+import os
+
+env=gym.make("CartPole-v0")
+env._max_episode_steps=200
+nbr_actions=2
+gamma=0.99
+max_episode=600
+
+prefix_log_file="log_actor"
+id_file=0
+while os.path.exists(prefix_log_file+str(id_file)+".csv"):
+    id_file+=1
+fichier_log=open(prefix_log_file+str(id_file)+".csv", "w")
+print("Création du fichier de log", prefix_log_file+str(id_file)+".csv")
+
+def model(nbr_inputs, nbr_hidden, nbr_actions):
+  entree=layers.Input(shape=(nbr_inputs), dtype='float32')
+  result=layers.Dense(32, activation='relu')(entree)
+  result=layers.Dense(32, activation='relu')(result)
+  sortie=layers.Dense(nbr_actions, activation='softmax')(result)
+
+  my_model=models.Model(inputs=entree, outputs=sortie)
+  return my_model
+
+def calcul_discount_rate(rewards_history, gamma, normalize=False):
+    result=[]
+    discounted_sum=0
+    for r in rewards_history[::-1]:
+        discounted_sum=r+gamma*discounted_sum
+        result.insert(0, discounted_sum)
+
+    # Normalisation
+    if normalize is True:
+        result=np.array(result)
+        result=(result-np.mean(result))/(np.std(result)+1E-7)
+        result=list(result)
+
+    return result
+
+def train():
+  m_reward=0
+  for episode in range(max_episode):
+    tab_rewards=[]
+    tab_prob_actions=[]
+
+    observations=env.reset()
+    with tf.GradientTape() as tape:
+        while True:
+                action_probs=my_model(np.expand_dims(observations, axis=0))
+                action=np.random.choice(nbr_actions, p=np.squeeze(action_probs))
+                tab_prob_actions.append(action_probs[0, action])
+                observations, reward, done, info=env.step(action)
+                tab_rewards.append(reward)
+                if done:
+                    break
+
+        discount_rate=calcul_discount_rate(tab_rewards, gamma, normalize=True)
+
+        loss=-tf.math.log(tab_prob_actions)*discount_rate
+        gradients=tape.gradient(loss, my_model.trainable_variables)
+        optimizer.apply_gradients(zip(gradients, my_model.trainable_variables))
+
+    score=sum(tab_rewards)
+    m_reward=0.05*score+(1-0.05)*m_reward
+    message="Episode {:04d}  score:{:6.1f}  MPE: {:6.1f}"
+    print(message.format(episode, score, m_reward))
+
+    fichier_log.write("{:f}:{:f}\n".format(score, m_reward))
+
+    if m_reward>env._max_episode_steps-10:
+        print("Fin de l'apprentissage".format(episode))
+        break
+
+my_model=model(4, 32, nbr_actions)
+optimizer=tf.keras.optimizers.Adam(learning_rate=1E-2)
+
+train()
+
+fichier_log.close()

Divers/renforcement6/cartpole_actor_critic.py

@@ -0,0 +1,103 @@
+import gym
+import numpy as np
+import tensorflow as tf
+from tensorflow import keras
+from tensorflow.keras import layers
+import os
+
+gamma=0.99
+max_steps_per_episode=10000
+env=gym.make("CartPole-v0")
+env._max_episode_steps=200
+
+prefix_log_file="log_actor_critic_dsum_"
+id_file=0
+while os.path.exists(prefix_log_file+str(id_file)+".csv"):
+    id_file+=1
+fichier_log=open(prefix_log_file+str(id_file)+".csv", "w")
+print("Création du fichier de log", prefix_log_file+str(id_file)+".csv")
+
+nbr_actions=2
+nbr_inputs=4
+
+def calcul_discount_rate(rewards_history, gamma, normalize=False):
+    result=[]
+    discounted_sum=0
+    for r in rewards_history[::-1]:
+        discounted_sum=r+gamma*discounted_sum
+        result.insert(0, discounted_sum)
+
+    # Normalisation
+    if normalize is True:
+        result=np.array(result)
+        result=(result-np.mean(result))/(np.std(result)+1E-7)
+        result=list(result)
+
+    return result
+
+def my_model(nbr_inputs, nbr_hidden, nbr_actions):
+    entree=layers.Input(shape=(nbr_inputs), dtype='float32')
+
+    common=layers.Dense(nbr_hidden, activation="relu")(entree)
+    action=layers.Dense(nbr_actions, activation="softmax")(common)
+    critic=layers.Dense(1)(common)
+
+    model=keras.Model(inputs=entree, outputs=[action, critic])
+    return model
+
+model=my_model(nbr_inputs, 32, nbr_actions)
+
+optimizer=keras.optimizers.Adam(learning_rate=1E-2)
+huber_loss=keras.losses.Huber()
+
+m_reward=0
+episode=0
+
+while True:
+    action_probs_history=[]
+    critic_value_history=[]
+    rewards_history=[]
+
+    state=env.reset()
+    episode_reward=0
+    with tf.GradientTape() as tape:
+
+        # Récupération de données
+        for timestep in range(1, max_steps_per_episode):
+            action_probs, critic_value=model(np.expand_dims(state, axis=0))
+            critic_value_history.append(critic_value[0, 0])
+            action=np.random.choice(nbr_actions, p=np.squeeze(action_probs))
+            action_probs_history.append(action_probs[0, action])
+            state, reward, done, infos=env.step(action)
+            rewards_history.append(reward)
+            episode_reward+=reward
+            if done:
+                break
+
+        discount_rate=calcul_discount_rate(rewards_history, gamma, normalize=True)
+
+        history=zip(action_probs_history, critic_value_history, discount_rate)
+        actor_losses=[]
+        critic_losses=[]
+        for action_prob, critic_value, discount_rate in history:
+            actor_losses.append(-tf.math.log(action_prob)*(discount_rate-critic_value))
+            critic_losses.append(huber_loss([critic_value], [discount_rate]))
+
+        loss_value=tf.reduce_mean(actor_losses+critic_losses)
+        grads=tape.gradient(loss_value, model.trainable_variables)
+        optimizer.apply_gradients(zip(grads, model.trainable_variables))
+
+    episode+=1
+    m_reward=0.05*episode_reward+(1-0.05)*m_reward
+
+    message="Episode {:04d}  score:{:6.1f}  MPE: {:6.1f}"
+    print(message.format(episode, episode_reward, m_reward))
+
+    fichier_log.write("{:f}:{:f}\n".format(episode_reward, m_reward))
+
+    if m_reward>env._max_episode_steps-10:
+        print("Fin de l'apprentissage".format(episode))
+        break
+
+fichier_log.close()
+model.save("my_model")

Divers/renforcement6/cartpole_critic.py

@@ -0,0 +1,114 @@
+import gym
+import tensorflow as tf
+from tensorflow.keras import models, layers
+import numpy as np
+import os
+
+env=gym.make("CartPole-v0")
+env._max_episode_steps=200
+nbr_action=2
+
+prefix_log_file="log_critic_"
+id_file=0
+while os.path.exists(prefix_log_file+str(id_file)+".csv"):
+    id_file+=1
+fichier_log=open(prefix_log_file+str(id_file)+".csv", "w")
+print("Création du fichier de log", prefix_log_file+str(id_file)+".csv")
+
+gamma=0.98
+max_episode=600
+epsilon=1.
+epsilon_min=0.10
+start_epsilon=10
+end_epsilon=max_episode
+epsilon_decay_value=epsilon/(end_epsilon-start_epsilon)
+
+def model():
+  entree=layers.Input(shape=(4), dtype='float32')
+  result=layers.Dense(32, activation='relu')(entree)
+  result=layers.Dense(32, activation='relu')(result)
+  sortie=layers.Dense(nbr_action)(result)
+
+  model=models.Model(inputs=entree, outputs=sortie)
+  return model
+
+def my_loss(target_q, predicted_q):
+  loss=tf.reduce_mean(tf.math.square(target_q-predicted_q))
+  return loss
+
+@tf.function
+def train_step(reward, action, observation, next_observation, done):
+  next_Q_values=model(next_observation)
+  best_next_actions=tf.math.argmax(next_Q_values, axis=1)
+  next_mask=tf.one_hot(best_next_actions, nbr_action)
+  next_best_Q_values=tf.reduce_sum(next_Q_values*next_mask, axis=1)
+  target_Q_values=reward+(1-done)*gamma*next_best_Q_values
+  target_Q_values=tf.reshape(target_Q_values, (-1, 1))
+  mask=tf.one_hot(action, nbr_action)
+  with tf.GradientTape() as tape:
+    all_Q_values=model(observation)
+    Q_values=tf.reduce_sum(all_Q_values*mask, axis=1, keepdims=True)
+    loss=my_loss(target_Q_values, Q_values)
+  gradients=tape.gradient(loss, model.trainable_variables)
+  optimizer.apply_gradients(zip(gradients, model.trainable_variables))
+  train_loss(loss)
+
+def train(debug=False):
+  global epsilon
+  m_reward=0
+  for episode in range(max_episode):
+    score=0
+    tab_observations=[]
+    tab_rewards=[]
+    tab_actions=[]
+    tab_next_observations=[]
+    tab_done=[]
+
+    observations=env.reset()
+    score=0
+    while True:
+      tab_observations.append(observations)
+      if np.random.random()>epsilon:
+        valeurs_q=model(np.expand_dims(observations, axis=0))
+        action=int(tf.argmax(valeurs_q[0], axis=-1))
+      else:
+        action=np.random.randint(0, nbr_action)
+      observations, reward, done, info=env.step(action)
+      score+=reward
+      tab_actions.append(action)
+      tab_next_observations.append(observations)
+      tab_done.append(done)
+      if done:
+        tab_rewards.append(-10.)
+        break
+      tab_rewards.append(reward)
+
+    tab_rewards=np.array(tab_rewards, dtype=np.float32)
+    tab_actions=np.array(tab_actions, dtype=np.int32)
+    tab_observations=np.array(tab_observations, dtype=np.float32)
+    tab_next_observations=np.array(tab_next_observations, dtype=np.float32)
+    tab_done=np.array(tab_done, dtype=np.float32)
+    train_step(tab_rewards, tab_actions, tab_observations, tab_next_observations, tab_done)
+    train_loss.reset_states()
+
+    epsilon-=epsilon_decay_value
+    epsilon=max(epsilon, epsilon_min)
+
+    m_reward=0.05*score+(1-0.05)*m_reward
+    message="Episode {:04d}  score:{:6.1f}  MPE: {:6.1f} (epsilon={:5.3f})"
+    print(message.format(episode, score, m_reward, epsilon))
+
+    fichier_log.write("{:f}:{:f}\n".format(score, m_reward))
+
+    if m_reward>env._max_episode_steps-10:
+        print("Fin de l'apprentissage".format(episode))
+        break
+
+model=model()
+optimizer=tf.keras.optimizers.Adam(learning_rate=1E-2)
+train_loss=tf.keras.metrics.Mean()
+tab_s=[]
+
+train()
+
+fichier_log.close()

Divers/tutoriel11/README.md

@@ -0,0 +1,4 @@
+# Tutoriel 11
+## Utilisation de Labelimg et conversion des fichiers xml en np.array
+
+La vidéo de ce tutoriel est disponible à l'adresse suivante: https://www.youtube.com/watch?v=VWXXFFDqBqA

Divers/tutoriel11/xmltoarray.py

@@ -0,0 +1,39 @@
+import xmltodict
+import numpy as np
+import glob
+import cv2
+
+def xml_to_dataset(dir, size=None):
+    tab_image=[]
+    tab_label=[]
+    for fichier in glob.glob(dir+"/*.xml"):
+        with open(fichier) as fd:
+            doc=xmltodict.parse(fd.read())
+            image=doc['annotation']['filename']
+            img=cv2.imread(dir+"/"+image)
+            objects=doc['annotation']['object'] if type(doc['annotation']['object'])==list else [doc['annotation']['object']]
+            for obj in objects:
+                xmin=int(obj['bndbox']['xmin'])
+                xmax=int(obj['bndbox']['xmax'])
+                ymin=int(obj['bndbox']['ymin'])
+                ymax=int(obj['bndbox']['ymax'])
+                if size is not None:
+                    tab_image.append(cv2.resize(img[ymin:ymax, xmin:xmax], size))
+                else:
+                    tab_image.append(img[ymin:ymax, xmin:xmax])
+                tab_label.append(obj['name'])
+
+    l=list(set(tab_label))
+    tab_one_hot=[]
+    for e in tab_label:
+        tab_one_hot.append(np.eye(len(l))[l.index(e)])
+        
+    return tab_image, l, tab_one_hot
+
+tab_image, tab_label, tab_one_hot=xml_to_dataset("./", (32, 32))
+
+for i in range(len(tab_image)):
+    cv2.imshow('image', tab_image[i])
+    print(tab_label[np.argmax(tab_one_hot[i])], tab_one_hot[i])
+    cv2.waitKey()
+

Divers/tutoriel12/README.md

@@ -0,0 +1,6 @@
+# Tutoriel 12
+## Utilisation basique de réseau de neurones dédié à la vision (modèle zoo)
+
+La vidéo de ce tutoriel est disponible à l'adresse suivante: https://www.youtube.com/watch?v=r2U-ntB-RM4
+
+Depuis la vidéo, la version tensorflow 2 est sortie, le programme 'detection_tf2.py' est la version en tensorflow 2 (testé avec la version 2.3 cpu)

Divers/tutoriel12/detection.py

@@ -0,0 +1,146 @@
+import numpy as np
+import os
+import tensorflow as tf
+import cv2
+
+labels={
+ 1: "person",
+ 2: "bicycle",
+ 3: "car",
+ 4: "motorcycle",
+ 5: "airplane",
+ 6: "bus",
+ 7: "train",
+ 8: "truck",
+ 9: "boat",
+ 10: "traffic light",
+ 11: "fire hydrant",
+ 13: "stop sign",
+ 14: "parking meter",
+ 15: "bench",
+ 16: "bird",
+ 17: "cat",
+ 18: "dog",
+ 19: "horse",
+ 20: "sheep",
+ 21: "cow",
+ 22: "elephant",
+ 23: "bear",
+ 24: "zebra",
+ 25: "giraffe",
+ 27: "backpack",
+ 28: "umbrella",
+ 31: "handbag",
+ 32: "tie",
+ 33: "suitcase",
+ 34: "frisbee",
+ 35: "skis",
+ 36: "snowboard",
+ 37: "sports ball",
+ 38: "kite",
+ 39: "baseball bat",
+ 40: "baseball glove",
+ 41: "skateboard",
+ 42: "surfboard",
+ 43: "tennis racket",
+ 44: "bottle",
+ 46: "wine glass",
+ 47: "cup",
+ 48: "fork",
+ 49: "knife",
+ 50: "spoon",
+ 51: "bowl",
+ 52: "banana",
+ 53: "apple",
+ 54: "sandwich",
+ 55: "orange",
+ 56: "broccoli",
+ 57: "carrot",
+ 58: "hot dog",
+ 59: "pizza",
+ 60: "donut",
+ 61: "cake",
+ 62: "chair",
+ 63: "couch",
+ 64: "potted plant",
+ 65: "bed",
+ 67: "dining table",
+ 70: "toilet",
+ 72: "tv",
+ 73: "laptop",
+ 74: "mouse",
+ 75: "remote",
+ 76: "keyboard",
+ 77: "cell phone",
+ 78: "microwave",
+ 79: "oven",
+ 80: "toaster",
+ 81: "sink",
+ 82: "refrigerator",
+ 84: "book",
+ 85: "clock",
+ 86: "vase",
+ 87: "scissors",
+ 88: "teddy bear",
+ 89: "hair drier",
+ 90: "toothbrush"
+}
+
+MODEL_NAME='ssd_mobilenet_v2_coco_2018_03_29'
+PATH_TO_FROZEN_GRAPH=MODEL_NAME+'/frozen_inference_graph.pb'
+color_infos=(255, 255, 0)
+
+detection_graph=tf.Graph()
+with detection_graph.as_default():
+  od_graph_def=tf.GraphDef()
+  with tf.gfile.GFile(PATH_TO_FROZEN_GRAPH, 'rb') as fid:
+    serialized_graph=fid.read()
+    od_graph_def.ParseFromString(serialized_graph)
+    tf.import_graph_def(od_graph_def, name='')
+
+with detection_graph.as_default():
+    with tf.Session() as sess:
+        cap=cv2.VideoCapture(0)
+        ops=tf.get_default_graph().get_operations()
+        all_tensor_names={output.name for op in ops for output in op.outputs}
+        tensor_dict={}
+        for key in [
+            'num_detections', 'detection_boxes', 'detection_scores',
+            'detection_classes', 'detection_masks']:
+            tensor_name=key+':0'
+            if tensor_name in all_tensor_names:
+                tensor_dict[key]=tf.get_default_graph().get_tensor_by_name(tensor_name)
+        if 'detection_masks' in tensor_dict:
+            quit("Masque non géré")
+        image_tensor=tf.get_default_graph().get_tensor_by_name('image_tensor:0')
+
+        while True:
+            ret, frame=cap.read()
+            tickmark=cv2.getTickCount()
+            output_dict=sess.run(tensor_dict, feed_dict={image_tensor: np.expand_dims(frame, 0)})
+            nbr_object=int(output_dict['num_detections'])
+            classes=output_dict['detection_classes'][0].astype(np.uint8)
+            boxes=output_dict['detection_boxes'][0]
+            scores=output_dict['detection_scores'][0]
+            for objet in range(nbr_object):
+                ymin, xmin, ymax, xmax=boxes[objet]
+                if scores[objet]>0.30:
+                    height, width=frame.shape[:2]
+                    xmin=int(xmin*width)
+                    xmax=int(xmax*width)
+                    ymin=int(ymin*height)
+                    ymax=int(ymax*height)
+                    cv2.rectangle(frame, (xmin, ymin), (xmax, ymax), color_infos, 1)
+                    txt="{:s}:{:3.0%}".format(labels[classes[objet]], scores[objet])
+                    cv2.putText(frame, txt, (xmin, ymin-5), cv2.FONT_HERSHEY_PLAIN, 1, color_infos, 2)
+            fps=cv2.getTickFrequency()/(cv2.getTickCount()-tickmark)
+            cv2.putText(frame, "FPS: {:05.2f}".format(fps), (10, 20), cv2.FONT_HERSHEY_PLAIN, 1, color_infos, 2)
+            cv2.imshow('image', frame)
+            key=cv2.waitKey(1)&0xFF
+            if key==ord('a'):
+                for objet in range(500):
+                    ret, frame=cap.read()
+            if key==ord('q'):
+              break
+cap.release()
+cv2.destroyAllWindows()

Divers/tutoriel12/detection_tf2.py

@@ -0,0 +1,148 @@
+import numpy as np
+import os
+import tensorflow as tf
+import cv2
+
+labels={
+ 1: "person",
+ 2: "bicycle",
+ 3: "car",
+ 4: "motorcycle",
+ 5: "airplane",
+ 6: "bus",
+ 7: "train",
+ 8: "truck",
+ 9: "boat",
+ 10: "traffic light",
+ 11: "fire hydrant",
+ 13: "stop sign",
+ 14: "parking meter",
+ 15: "bench",
+ 16: "bird",
+ 17: "cat",
+ 18: "dog",
+ 19: "horse",
+ 20: "sheep",
+ 21: "cow",
+ 22: "elephant",
+ 23: "bear",
+ 24: "zebra",
+ 25: "giraffe",
+ 27: "backpack",
+ 28: "umbrella",
+ 31: "handbag",
+ 32: "tie",
+ 33: "suitcase",
+ 34: "frisbee",
+ 35: "skis",
+ 36: "snowboard",
+ 37: "sports ball",
+ 38: "kite",
+ 39: "baseball bat",
+ 40: "baseball glove",
+ 41: "skateboard",
+ 42: "surfboard",
+ 43: "tennis racket",
+ 44: "bottle",
+ 46: "wine glass",
+ 47: "cup",
+ 48: "fork",
+ 49: "knife",
+ 50: "spoon",
+ 51: "bowl",
+ 52: "banana",
+ 53: "apple",
+ 54: "sandwich",
+ 55: "orange",
+ 56: "broccoli",
+ 57: "carrot",
+ 58: "hot dog",
+ 59: "pizza",
+ 60: "donut",
+ 61: "cake",
+ 62: "chair",
+ 63: "couch",
+ 64: "potted plant",
+ 65: "bed",
+ 67: "dining table",
+ 70: "toilet",
+ 72: "tv",
+ 73: "laptop",
+ 74: "mouse",
+ 75: "remote",
+ 76: "keyboard",
+ 77: "cell phone",
+ 78: "microwave",
+ 79: "oven",
+ 80: "toaster",
+ 81: "sink",
+ 82: "refrigerator",
+ 84: "book",
+ 85: "clock",
+ 86: "vase",
+ 87: "scissors",
+ 88: "teddy bear",
+ 89: "hair drier",
+ 90: "toothbrush"
+}
+
+#MODEL_NAME='ssd_mobilenet_v2_coco_2018_03_29'
+MODEL_NAME='faster_rcnn_inception_v2_coco_2018_01_28'
+PATH_TO_FROZEN_GRAPH=MODEL_NAME+'/frozen_inference_graph.pb'
+color_infos=(255, 255, 0)
+
+detection_graph=tf.compat.v1.Graph()
+with detection_graph.as_default():
+  od_graph_def=tf.compat.v1.GraphDef()
+  with tf.io.gfile.GFile(PATH_TO_FROZEN_GRAPH, 'rb') as fid:
+    serialized_graph=fid.read()
+    od_graph_def.ParseFromString(serialized_graph)
+    tf.import_graph_def(od_graph_def, name='')
+
+with detection_graph.as_default():
+    with tf.compat.v1.Session() as sess:
+        #cap=cv2.VideoCapture('../../En_cours/tuto8-suite/Plan_9_from_Outer_Space_1959_512kb.mp4')
+        cap=cv2.VideoCapture(0)
+        ops=tf.compat.v1.get_default_graph().get_operations()
+        all_tensor_names={output.name for op in ops for output in op.outputs}
+        tensor_dict={}
+        for key in [
+            'num_detections', 'detection_boxes', 'detection_scores',
+            'detection_classes', 'detection_masks']:
+            tensor_name=key+':0'
+            if tensor_name in all_tensor_names:
+                tensor_dict[key]=tf.compat.v1.get_default_graph().get_tensor_by_name(tensor_name)
+        if 'detection_masks' in tensor_dict:
+            quit("Masque non géré")
+        image_tensor=tf.compat.v1.get_default_graph().get_tensor_by_name('image_tensor:0')
+
+        while True:
+            ret, frame=cap.read()
+            tickmark=cv2.getTickCount()
+            output_dict=sess.run(tensor_dict, feed_dict={image_tensor: np.expand_dims(frame, 0)})
+            nbr_object=int(output_dict['num_detections'])
+            classes=output_dict['detection_classes'][0].astype(np.uint8)
+            boxes=output_dict['detection_boxes'][0]
+            scores=output_dict['detection_scores'][0]
+            for objet in range(nbr_object):
+                ymin, xmin, ymax, xmax=boxes[objet]
+                if scores[objet]>0.30:
+                    height, width=frame.shape[:2]
+                    xmin=int(xmin*width)
+                    xmax=int(xmax*width)
+                    ymin=int(ymin*height)
+                    ymax=int(ymax*height)
+                    cv2.rectangle(frame, (xmin, ymin), (xmax, ymax), color_infos, 1)
+                    txt="{:s}:{:3.0%}".format(labels[classes[objet]], scores[objet])
+                    cv2.putText(frame, txt, (xmin, ymin-5), cv2.FONT_HERSHEY_PLAIN, 1, color_infos, 2)
+            fps=cv2.getTickFrequency()/(cv2.getTickCount()-tickmark)
+            cv2.putText(frame, "FPS: {:05.2f}".format(fps), (10, 20), cv2.FONT_HERSHEY_PLAIN, 1, color_infos, 2)
+            cv2.imshow('image', frame)
+            key=cv2.waitKey(1)&0xFF
+            if key==ord('a'):
+                for objet in range(500):
+                    ret, frame=cap.read()
+            if key==ord('q'):
+              break
+cap.release()
+cv2.destroyAllWindows()

Divers/tutoriel18-1/README.md

@@ -0,0 +1,5 @@
+# Tutoriel 18 1artie 1
+## Sudoku
+
+La vidéo de ce tutoriel est disponible à l'adresse suivante: https://www.youtube.com/watch?v=WwPHs1SJrec
+

Divers/tutoriel18-1/perspective.py

@@ -0,0 +1,69 @@
+import cv2
+import numpy as np
+import operator
+
+marge=4
+case=28+2*marge
+taille_grille=9*case
+
+methode=cv2.ADAPTIVE_THRESH_GAUSSIAN_C
+v1=9
+
+cap=cv2.VideoCapture(0)
+while True:
+    ret, frame=cap.read()
+    gray=cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
+    gray=cv2.GaussianBlur(gray, (5, 5), 0)
+    thresh=cv2.adaptiveThreshold(gray, 255, methode, cv2.THRESH_BINARY_INV, v1, 2)
+    #cv2.imshow("thresh", thresh)
+    contours, hierarchy=cv2.findContours(thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
+    contour_grille=None
+    maxArea=0
+    for c in contours:
+        area=cv2.contourArea(c)
+        if area>25000:
+            peri=cv2.arcLength(c, True)
+            polygone=cv2.approxPolyDP(c, 0.01*peri, True)
+            if area>maxArea and len(polygone)==4:
+                contour_grille=polygone
+                maxArea=area
+    if contour_grille is not None:
+        cv2.drawContours(frame, [contour_grille], 0, (0, 255, 0), 2)
+        points=np.vstack(contour_grille).squeeze()
+        points=sorted(points, key=operator.itemgetter(1))
+        if points[0][0]<points[1][0]:
+            if points[3][0]<points[2][0]:
+                pts1=np.float32([points[0], points[1], points[3], points[2]])
+            else:
+                pts1=np.float32([points[0], points[1], points[2], points[3]])
+        else:
+            if points[3][0]<points[2][0]:
+                pts1=np.float32([points[1], points[0], points[3], points[2]])
+            else:
+                pts1=np.float32([points[1], points[0], points[2], points[3]])
+        pts2=np.float32([[0, 0], [taille_grille, 0], [0, taille_grille], [taille_grille, taille_grille]])
+        M=cv2.getPerspectiveTransform(pts1, pts2)
+        grille=cv2.warpPerspective(frame, M, (taille_grille, taille_grille))
+        cv2.putText(frame, "1", (points[0][0], points[0][1]), cv2.FONT_HERSHEY_COMPLEX_SMALL, 0.9, (0, 0, 255), 1)
+        cv2.putText(frame, "2", (points[1][0], points[1][1]), cv2.FONT_HERSHEY_COMPLEX_SMALL, 0.9, (0, 0, 255), 1)
+        cv2.putText(frame, "3", (points[2][0], points[2][1]), cv2.FONT_HERSHEY_COMPLEX_SMALL, 0.9, (0, 0, 255), 1)
+        cv2.putText(frame, "4", (points[3][0], points[3][1]), cv2.FONT_HERSHEY_COMPLEX_SMALL, 0.9, (0, 0, 255), 1)
+        cv2.imshow("grille", grille)
+    txt="ADAPTIVE_THRESH_MEAN_C" if methode==cv2.ADAPTIVE_THRESH_MEAN_C else "ADAPTIVE_THRESH_GAUSSIAN_C"
+    cv2.putText(frame, "[p|m]v1: {:2d}  [o]methode: {}".format(v1, txt), (10, 20), cv2.FONT_HERSHEY_COMPLEX_SMALL, 0.9, (0, 0, 255), 1)
+    cv2.imshow("frame", frame)
+    key=cv2.waitKey(1)&0xFF
+    if key==ord('q'):
+        break
+    if key==ord('p'):
+        v1=min(21, v1+2)
+    if key==ord('m'):
+        v1=max(3, v1-2)
+        print(v1)
+    if key==ord('o'):
+        if methode==cv2.ADAPTIVE_THRESH_GAUSSIAN_C:
+            methode=cv2.ADAPTIVE_THRESH_MEAN_C
+        else:
+            methode=cv2.ADAPTIVE_THRESH_GAUSSIAN_C
+cap.release()
+cv2.destroyAllWindows()

Divers/tutoriel18-1/poly.py

@@ -0,0 +1,47 @@
+import cv2
+import numpy as np
+import operator
+
+methode=cv2.ADAPTIVE_THRESH_GAUSSIAN_C
+v1=9
+
+cap=cv2.VideoCapture(0)
+while True:
+    ret, frame=cap.read()
+    gray=cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
+    gray=cv2.GaussianBlur(gray, (5, 5), 0)
+    thresh=cv2.adaptiveThreshold(gray, 255, methode, cv2.THRESH_BINARY_INV, v1, 2)
+    cv2.imshow("thresh", thresh)
+
+    contours, hierarchy=cv2.findContours(thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
+    contour_grille=None
+    maxArea=0
+    for c in contours:
+        area=cv2.contourArea(c)
+        if area>25000:
+            peri=cv2.arcLength(c, True)
+            polygone=cv2.approxPolyDP(c, 0.01*peri, True)
+            if area>maxArea and len(polygone)==4:
+                contour_grille=polygone
+                maxArea=area
+    if contour_grille is not None:
+        cv2.drawContours(frame, [contour_grille], 0, (0, 255, 0), 2)
+    txt="ADAPTIVE_THRESH_MEAN_C" if methode==cv2.ADAPTIVE_THRESH_MEAN_C else "ADAPTIVE_THRESH_GAUSSIAN_C"
+    cv2.putText(frame, "[p|m]v1: {:2d}  [o]methode: {}".format(v1, txt), (10, 20), cv2.FONT_HERSHEY_COMPLEX_SMALL, 0.9, (0, 0, 255), 1)
+
+    cv2.imshow("frame", frame)
+    key=cv2.waitKey(1)&0xFF
+    if key==ord('q'):
+        break
+    if key==ord('p'):
+        v1=min(21, v1+2)
+    if key==ord('m'):
+        v1=max(3, v1-2)
+        print(v1)
+    if key==ord('o'):
+        if methode==cv2.ADAPTIVE_THRESH_GAUSSIAN_C:
+            methode=cv2.ADAPTIVE_THRESH_MEAN_C
+        else:
+            methode=cv2.ADAPTIVE_THRESH_GAUSSIAN_C
+cap.release()
+cv2.destroyAllWindows()

Divers/tutoriel18-1/threshold.py

@@ -0,0 +1,31 @@
+import cv2
+import numpy as np
+import operator
+
+methode=cv2.ADAPTIVE_THRESH_GAUSSIAN_C
+v1=9
+
+cap=cv2.VideoCapture(0)
+while True:
+    ret, frame=cap.read()
+    gray=cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
+    gray=cv2.GaussianBlur(gray, (5, 5), 0)
+    thresh=cv2.adaptiveThreshold(gray, 255, methode, cv2.THRESH_BINARY_INV, v1, 2)
+    cv2.imshow("thresh", thresh)
+    txt="ADAPTIVE_THRESH_MEAN_C" if methode==cv2.ADAPTIVE_THRESH_MEAN_C else "ADAPTIVE_THRESH_GAUSSIAN_C"
+    cv2.putText(frame, "[p|m]v1: {:2d}  [o]methode: {}".format(v1, txt), (10, 20), cv2.FONT_HERSHEY_COMPLEX_SMALL, 0.9, (0, 0, 255), 1)
+    cv2.imshow("frame", frame)
+    key=cv2.waitKey(1)&0xFF
+    if key==ord('q'):
+        break
+    if key==ord('p'):
+        v1=min(21, v1+2)
+    if key==ord('m'):
+        v1=max(3, v1-2)
+    if key==ord('o'):
+        if methode==cv2.ADAPTIVE_THRESH_GAUSSIAN_C:
+            methode=cv2.ADAPTIVE_THRESH_MEAN_C
+        else:
+            methode=cv2.ADAPTIVE_THRESH_GAUSSIAN_C
+cap.release()
+cv2.destroyAllWindows()

Divers/tutoriel18-2/README.md

@@ -0,0 +1,5 @@
+# Tutoriel 18 partie 2
+## Sudoku 
+
+La vidéo de ce tutoriel est disponible à l'adresse suivante: https://www.youtube.com/watch?v=XFNg8lXe-Tk
+

Divers/tutoriel18-2/font.py

@@ -0,0 +1,18 @@
+from PIL import ImageFont, ImageDraw, Image
+import cv2
+import numpy as np
+
+for i in range(1, 10):                
+    image=Image.new("L", (28, 28))
+    draw=ImageDraw.Draw(image)
+    font=ImageFont.truetype("din1451altG.ttf", 27)
+    text="{:d}".format(i)
+    draw.text((10, 0), text, font=font, fill=(255))
+    image=np.array(image).reshape(28, 28, 1)
+    cv2.imshow("image", image)
+    key=cv2.waitKey()
+    if key&0xFF==ord('q'):
+        quit()
+                                                                    
+
+    

Divers/tutoriel18-2/sudoku.py

@@ -0,0 +1,100 @@
+import cv2
+import numpy as np
+import tensorflow as tf
+import sudoku_solver as ss
+from time import sleep
+import operator
+
+marge=4
+case=28+2*marge
+taille_grille=9*case
+flag=0
+cap=cv2.VideoCapture(0)
+with tf.Session() as s:
+    saver=tf.train.import_meta_graph('./mon_modele/modele.meta')
+    saver.restore(s, tf.train.latest_checkpoint('./mon_modele/'))
+    graph=tf.get_default_graph()
+    images=graph.get_tensor_by_name("entree:0")
+    sortie=graph.get_tensor_by_name("sortie:0")
+    is_training=graph.get_tensor_by_name("is_training:0")
+    maxArea=0
+    while True:
+        ret, frame=cap.read()
+        if maxArea==0:
+            cv2.imshow("frame", frame)
+        gray=cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
+        gray=cv2.GaussianBlur(gray, (5, 5), 0)
+        thresh=cv2.adaptiveThreshold(gray, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY_INV, 9, 2)
+        contours, hierarchy=cv2.findContours(thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
+        contour_grille=None
+        maxArea=0
+        for c in contours:
+            area=cv2.contourArea(c)
+            if area>25000:
+                peri=cv2.arcLength(c, True)
+                polygone=cv2.approxPolyDP(c, 0.02*peri, True)
+                if area>maxArea and len(polygone)==4:
+                    contour_grille=polygone
+                    maxArea=area
+        if contour_grille is not None:
+            points=np.vstack(contour_grille).squeeze()
+            points=sorted(points, key=operator.itemgetter(1))
+            if points[0][0]<points[1][0]:
+                if points[3][0]<points[2][0]:
+                    pts1=np.float32([points[0], points[1], points[3], points[2]])
+                else:
+                    pts1=np.float32([points[0], points[1], points[2], points[3]])
+            else:
+                if points[3][0]<points[2][0]:
+                    pts1=np.float32([points[1], points[0], points[3], points[2]])
+                else:
+                    pts1=np.float32([points[1], points[0], points[2], points[3]])
+            pts2=np.float32([[0, 0], [taille_grille, 0], [0, taille_grille], [taille_grille, taille_grille]])
+            M=cv2.getPerspectiveTransform(pts1, pts2)
+            grille=cv2.warpPerspective(frame, M, (taille_grille, taille_grille))
+            grille=cv2.cvtColor(grille, cv2.COLOR_BGR2GRAY)
+            grille=cv2.adaptiveThreshold(grille, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY_INV, 9, 2)
+            cv2.imshow("grille", grille)
+            if flag==0:
+                grille=grille/255
+                grille_txt=[]
+                for y in range(9):
+                    ligne=""
+                    for x in range(9):
+                        y2min=y*case+marge
+                        y2max=(y+1)*case-marge
+                        x2min=x*case+marge
+                        x2max=(x+1)*case-marge
+                        prediction=s.run(sortie, feed_dict={images: [grille[y2min:y2max, x2min:x2max].reshape(28, 28, 1)], is_training: False})
+                        ligne+="{:d}".format(np.argmax(prediction[0]))
+                    grille_txt.append(ligne)
+                result=ss.sudoku(grille_txt)
+                print("Resultat:", result)
+            #result=None
+            if result is not None:
+                flag=1
+                fond=np.zeros(shape=(taille_grille, taille_grille, 3), dtype=np.float32)
+                for y in range(len(result)):
+                    for x in range(len(result[y])):
+                            if grille_txt[y][x]=="0":
+                                cv2.putText(fond, "{:d}".format(result[y][x]), ((x)*case+marge+3, (y+1)*case-marge-3), cv2.FONT_HERSHEY_SCRIPT_COMPLEX, 0.9, (0, 0, 255), 1)
+                M=cv2.getPerspectiveTransform(pts2, pts1)
+                h, w, c=frame.shape
+                fondP=cv2.warpPerspective(fond, M, (w, h))
+                img2gray=cv2.cvtColor(fondP, cv2.COLOR_BGR2GRAY)
+                ret, mask=cv2.threshold(img2gray, 10, 255, cv2.THRESH_BINARY)
+                mask=mask.astype('uint8')
+                mask_inv=cv2.bitwise_not(mask)
+                img1_bg=cv2.bitwise_and(frame, frame, mask=mask_inv)
+                img2_fg=cv2.bitwise_and(fondP, fondP, mask=mask).astype('uint8')
+                dst=cv2.add(img1_bg, img2_fg)
+                cv2.imshow("frame", dst)
+            else:
+                cv2.imshow("frame", frame)
+        else:
+            flag=0
+        key=cv2.waitKey(1)&0xFF
+        if key==ord('q'):
+            break
+cap.release()
+cv2.destroyAllWindows()

Divers/tutoriel18-2/sudoku_solver.py

@@ -0,0 +1,89 @@
+def sudoku(f):
+
+    def af(g):
+      for n,l in enumerate(g):
+           for m,c in enumerate(l):
+                P(str(c).replace("0","."),end="")
+                if m in {2,5}:
+                    P("+",end="")
+           P()
+           if n in {2,5}:
+               P("+"*11)
+
+    def cp(q,s):
+        l=set(s[q[0]])
+        l|={s[i][q[1]] for i in range(9)}
+        k=q[0]//3,q[1]//3
+        for i in range(3):
+            l|=set(s[k[0]*3+i][k[1]*3:(k[1]+1)*3])
+        return set(range(1,10))-l
+
+    def ec(l):
+        q=set(l)-{0}
+        for c in q:
+            if l.count(c)!=1:
+                return True
+        return False
+
+     # Remplissage de la grille et tests de format
+
+    P=print
+    af(f)
+
+    s=[]
+    t=[]
+    for nl,l in enumerate(f):
+        try:
+            n=list(map(int,l))
+        except:
+            P("La ligne "+str(nl+1)+" contient autre chose qu'un chiffre.")
+            return
+        if len(n)!=9:
+            P("La ligne "+str(nl+1)+" ne contient pas 9 chiffres.")
+            return
+        t+=[[nl,i] for i in range(9) if n[i]==0]
+        s.append(n)
+    if nl!=8:
+        P("Le jeu contient "+str(nl+1)+" lignes au lieu de 9.")
+        return
+     # Tests de validite
+
+    for l in range(9):
+        if ec(s[l]):
+           P("La ligne "+str(l+1)+" est contradictoire.")
+           return
+    for c in range(9):
+        k=[s[l][c] for l in range(9)]
+        if ec(k):
+           P("La colonne "+str(c+1)+" est contradictoire.")
+           return
+    for l in range(3):
+        for c in range(3):
+            q=[]
+            for i in range(3):
+                q+=s[l*3+i][c*3:(c+1)*3]
+            if ec(q):
+                P("La cellule ("+str(l+1)+";"+str(c+1)+") est contradictoire.")
+                return
+
+     # Resolution
+
+    p=[[] for i in t]
+    cr=0
+
+    while cr<len(t):
+        p[cr]=cp(t[cr],s)
+        try:
+            while not p[cr]:
+                s[t[cr][0]][t[cr][1]]=0
+                cr-=1
+        except:
+           P("Le sudoku n'a pas de solution.")
+           return
+        s[t[cr][0]][t[cr][1]]=p[cr].pop()
+        cr+=1
+
+     # Presentation de la grille resolue
+
+    af(s)
+    return(s)

Divers/tutoriel18-2/train.py

@@ -0,0 +1,147 @@
+import tensorflow as tf
+import numpy as np
+from sklearn.utils import shuffle
+from sklearn.model_selection import train_test_split
+import cv2
+import os
+import numpy as np
+from PIL import ImageFont, ImageDraw, Image
+
+taille_batch=100
+nbr_entrainement=100
+nbr=2*42
+
+def modif_image(image, seuil=1):
+    b=np.random.normal(0, 1, (28, 28))
+    a=image.copy()
+    a[b>seuil]=255
+    a[b<-seuil]=0
+    return a
+
+def convolution(input, taille_noyau, nbr_noyau, stride, b_norm=False, f_activation=None, training=False):
+    w_filtre=tf.Variable(tf.random.truncated_normal(shape=(taille_noyau, taille_noyau, int(input.get_shape()[-1]), nbr_noyau)))
+    b_filtre=np.zeros(nbr_noyau)
+    result=tf.nn.conv2d(input, w_filtre, strides=[1, stride, stride, 1], padding='SAME')+b_filtre
+    if b_norm is True:
+        result=tf.layers.batch_normalization(result, training=training)
+    if f_activation is not None:
+        result=f_activation(result)
+    return result
+        
+def fc(input, nbr_neurone, b_norm=False, f_activation=None, training=False):
+    w=tf.Variable(tf.random.truncated_normal(shape=(int(input.get_shape()[-1]), nbr_neurone), dtype=tf.float32))
+    b=tf.Variable(np.zeros(shape=(nbr_neurone)), dtype=tf.float32)
+    result=tf.matmul(input, w)+b
+    if b_norm is True:
+        result=tf.layers.batch_normalization(result, training=training)
+    if f_activation is not None:
+        result=f_activation(result)                        
+    return result
+
+def ia(nbr_classes, size, couche, learning_rate=1E-3):
+    ph_images=tf.placeholder(shape=(None, size, size, couche), dtype=tf.float32, name='entree')
+    ph_labels=tf.placeholder(shape=(None, nbr_classes), dtype=tf.float32)
+    ph_is_training=tf.placeholder_with_default(False, (), name='is_training')
+    
+    result=convolution(ph_images, 3, 64, 1, True, tf.nn.relu, ph_is_training)
+    result=tf.layers.dropout(result, 0.3, training=ph_is_training)
+    result=convolution(result, 3, 128, 1, True, tf.nn.relu, ph_is_training)
+    result=tf.layers.dropout(result, 0.4, training=ph_is_training)
+    result=tf.nn.max_pool(result, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
+
+    result=tf.contrib.layers.flatten(result)
+    
+    result=fc(result, 128, True, tf.nn.relu, ph_is_training)
+    result=tf.layers.dropout(result, 0.5, training=ph_is_training)
+    result=fc(result, nbr_classes)
+    socs=tf.nn.softmax(result, name="sortie")
+    
+    loss=tf.nn.softmax_cross_entropy_with_logits_v2(labels=ph_labels, logits=result)
+    extra_update_ops=tf.get_collection(tf.GraphKeys.UPDATE_OPS)
+    with tf.control_dependencies(extra_update_ops):
+        train=tf.train.AdamOptimizer(learning_rate).minimize(loss)
+    accuracy=tf.reduce_mean(tf.cast(tf.equal(tf.argmax(socs, 1), tf.argmax(ph_labels, 1)), tf.float32))
+
+    return ph_images, ph_labels, ph_is_training, socs, train, accuracy, tf.train.Saver()
+
+tab_images=[]
+tab_labels=[]
+
+for dir in ["/usr/share/fonts/truetype/ubuntu-font-family/", "/usr/share/fonts/truetype/freefont/"]:
+    for root, dirs, files in os.walk(dir):
+        for file in files:
+            if file.endswith("ttf"):
+                print(root+"/"+file)
+                for i in range(1, 10):
+                    for cpt in range(nbr):
+                        image=Image.new("L", (28, 28))
+                        draw=ImageDraw.Draw(image)
+                        font=ImageFont.truetype(root+"/"+file, np.random.randint(26, 32))
+                        text="{:d}".format(i)
+                        draw.text((np.random.randint(1, 10), np.random.randint(-4, 0)), text, font=font, fill=(255))
+                        image=np.array(image).reshape(28, 28, 1)
+                        tab_images.append(image)
+                        tab_labels.append(np.eye(10)[i])                        
+                        image_m=modif_image(image, 1.05+np.random.rand())
+                        tab_images.append(image_m)
+                        tab_labels.append(np.eye(10)[i])
+                image=np.zeros((28, 28, 1))
+                for cpt in range(3*nbr):
+                    image_m=modif_image(image, 1.05+np.random.rand())
+                    tab_images.append(image_m)
+                    tab_labels.append(np.eye(10)[0])
+                    
+tab_images=np.array(tab_images)
+tab_labels=np.array(tab_labels)
+
+tab_images=tab_images/255
+
+tab_images, tab_labels=shuffle(tab_images, tab_labels)
+
+if False: # Changer en True si vous voulez voir les images générées
+    for i in range(len(tab_images)):
+        cv2.imshow('chiffre', tab_images[i].reshape(28, 28, 1))
+        print(tab_labels[i], np.argmax(tab_labels[i]))
+        if cv2.waitKey()&0xFF==ord('q'):
+            break
+
+print("Nbr:", len(tab_images))
+
+train_images, test_images, train_labels, test_labels=train_test_split(tab_images, tab_labels, test_size=0.10)
+
+images, labels, is_training, sortie, train, accuracy, saver=ia(10, 28, 1)
+
+with tf.Session() as s:
+    s.run(tf.global_variables_initializer())
+    tab_train=[]
+    tab_test=[]
+    for id_entrainement in np.arange(nbr_entrainement):
+        print("> Entrainement", id_entrainement)
+        for batch in np.arange(0, len(train_images), taille_batch):
+            s.run(train, feed_dict={
+                images: train_images[batch:batch+taille_batch],
+                labels: train_labels[batch:batch+taille_batch],
+                is_training: True
+            })
+        print("  entrainement OK")
+        tab_accuracy_train=[]
+        for batch in np.arange(0, len(train_images), taille_batch):
+            p=s.run(accuracy, feed_dict={
+                images: train_images[batch:batch+taille_batch],
+                labels: train_labels[batch:batch+taille_batch],
+                is_training: True
+            })
+            tab_accuracy_train.append(p)
+        print("  train:", np.mean(tab_accuracy_train))
+        tab_accuracy_test=[]
+        for batch in np.arange(0, len(test_images), taille_batch):
+            p=s.run(accuracy, feed_dict={
+                images: test_images[batch:batch+taille_batch],
+                labels: test_labels[batch:batch+taille_batch],
+                is_training: True
+            })
+            tab_accuracy_test.append(p)
+        print("  test :", np.mean(tab_accuracy_test))
+        tab_train.append(1-np.mean(tab_accuracy_train))
+        tab_test.append(1-np.mean(tab_accuracy_test))
+    saver.save(s, './mon_modele/modele')

Divers/tutoriel20/README.md

@@ -0,0 +1,7 @@
+# Tutoriel 20 
+## Dlib: Detection et évaluation de la position de la tête
+
+La vidéo de ce tutoriel est disponible à l'adresse suivante: https://www.youtube.com/watch?v=ibuEFfpVWlU
+
+
+![image](https://github.com/L42Project/Tutoriels/blob/master/Divers/tutoriel20/dlib-landmark-mean.png)

Divers/tutoriel20/cube.py

@@ -0,0 +1,113 @@
+import cv2
+import numpy as np
+import dlib
+import math
+
+cap=cv2.VideoCapture(0)
+#cap=cv2.VideoCapture("debat.webm")
+
+detector=dlib.get_frontal_face_detector()
+predictor=dlib.shape_predictor("shape_predictor_68_face_landmarks.dat")
+
+def tr(c, o, coeff):
+    return(int((c-o)*coeff)+o)
+
+def cube(image, pt1, pt2, a1, a2, a3):
+    color=(0, 255, 0)
+    epaisseur=2
+    offset=1.6
+    offset2=2
+
+    d_eyes=math.sqrt(math.pow(landmarks.part(36).x-landmarks.part(45).x, 2)+math.pow(landmarks.part(36).y-landmarks.part(45).y, 2))
+
+    ox1=int((-(pt2.y-pt1.y)+pt2.x-pt1.x)/2)+pt1.x
+    oy1=int(((pt2.x-pt1.x+pt2.y)-pt1.y)/2)+pt1.y
+
+    cv2.line(image,
+        (tr(pt1.x, ox1, offset), tr(pt1.y, oy1, offset)),
+        (tr(pt2.x, ox1, offset), tr(pt2.y, oy1, offset)),
+        color, epaisseur)
+    cv2.line(image,
+        (tr(pt2.x, ox1, offset), tr(pt2.y, oy1, offset)),
+        (tr(-(pt2.y-pt1.y)+pt2.x, ox1, offset), tr(pt2.x-pt1.x+pt2.y, oy1, offset)),
+        color, epaisseur)
+    cv2.line(image,
+        (tr(-(pt2.y-pt1.y)+pt2.x, ox1, offset), tr(pt2.x-pt1.x+pt2.y, oy1, offset)),
+        (tr(-(pt2.y-pt1.y)+pt1.x, ox1, offset), tr(pt2.x-pt1.x+pt1.y, oy1, offset)),
+        color, epaisseur)
+    cv2.line(image,
+        (tr(-(pt2.y-pt1.y)+pt1.x, ox1, offset), tr(pt2.x-pt1.x+pt1.y, oy1, offset)),
+        (tr(pt1.x, ox1, offset), tr(pt1.y, oy1, offset)),
+        color, epaisseur)
+
+    ox2=int((-(pt2.y-pt1.y)+pt2.x-pt1.x)/2)+pt1.x+int(a2)
+    oy2=int(((pt2.x-pt1.x+pt2.y)-pt1.y)/2)+pt1.y+int(a3)
+
+    cv2.line(image,
+        (tr(pt1.x+a2, ox2, offset2), tr(pt1.y+a3, oy2, offset2)),
+        (tr(pt2.x+a2, ox2, offset2), tr(pt2.y+a3, oy2, offset2)),
+        color, epaisseur)
+    cv2.line(image,
+        (tr(pt2.x+a2, ox2, offset2), tr(pt2.y+a3, oy2, offset2)),
+        (tr(-(pt2.y-pt1.y)+pt2.x+a2, ox2, offset2), tr(pt2.x-pt1.x+pt2.y+a3, oy2, offset2)),
+        color, epaisseur)
+    cv2.line(image,
+        (tr(-(pt2.y-pt1.y)+pt2.x+a2, ox2, offset2), tr(pt2.x-pt1.x+pt2.y+a3, oy2, offset2)),
+        (tr(-(pt2.y-pt1.y)+pt1.x+a2, ox2, offset2), tr(pt2.x-pt1.x+pt1.y+a3, oy2, offset2)),
+        color, epaisseur)
+    cv2.line(image,
+        (tr(-(pt2.y-pt1.y)+pt1.x+a2, ox2, offset2), tr(pt2.x-pt1.x+pt1.y+a3, oy2, offset2)),
+        (tr(pt1.x+a2, ox2, offset2), tr(pt1.y+a3, oy2, offset2)),
+        color, epaisseur)
+
+    cv2.line(image,
+        (tr(pt1.x, ox1, offset), tr(pt1.y, oy1, offset)),
+        (tr(pt1.x+a2, ox2, offset2), tr(pt1.y+a3, oy2, offset2)),
+        color, epaisseur)
+    cv2.line(image,
+        (tr(pt2.x, ox1, offset), tr(pt2.y, oy1, offset)),
+        (tr(pt2.x+a2, ox2, offset2), tr(pt2.y+a3, oy2, offset2)),
+        color, epaisseur)
+    cv2.line(image,
+        (tr(-(pt2.y-pt1.y)+pt2.x, ox1, offset), tr(pt2.x-pt1.x+pt2.y, oy1, offset)),
+        (tr(-(pt2.y-pt1.y)+pt2.x+a2, ox2, offset2), tr(pt2.x-pt1.x+pt2.y+a3, oy2, offset2)),
+        color, epaisseur)
+    cv2.line(image,
+        (tr(-(pt2.y-pt1.y)+pt1.x, ox1, offset), tr(pt2.x-pt1.x+pt1.y, oy1, offset)),
+        (tr(-(pt2.y-pt1.y)+pt1.x+a2, ox2, offset2), tr(pt2.x-pt1.x+pt1.y+a3, oy2, offset2)),
+        color, epaisseur)
+
+while True:
+    ret, frame=cap.read()
+    tickmark=cv2.getTickCount()
+    gray=cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
+
+    faces=detector(gray)
+    for face in faces:
+        x1=face.left()
+        y1=face.top()
+        x2=face.right()
+        y2=face.bottom()
+
+        landmarks=predictor(gray, face)
+
+        d_eyes=math.sqrt(math.pow(landmarks.part(36).x-landmarks.part(45).x, 2)+math.pow(landmarks.part(36).y-landmarks.part(45).y, 2))
+        d1=math.sqrt(math.pow(landmarks.part(36).x-landmarks.part(30).x, 2)+math.pow(landmarks.part(36).y-landmarks.part(30).y, 2))
+        d2=math.sqrt(math.pow(landmarks.part(45).x-landmarks.part(30).x, 2)+math.pow(landmarks.part(45).y-landmarks.part(30).y, 2))
+        coeff=d1+d2
+
+        a1=int(250*(landmarks.part(36).y-landmarks.part(45).y)/coeff)
+        a2=int(250*(d1-d2)/coeff)
+        cosb=min((math.pow(d2, 2)-math.pow(d1, 2)+math.pow(d_eyes, 2))/(2*d2*d_eyes), 1)
+        a3=int(250*(d2*math.sin(math.acos(cosb))-coeff/4)/coeff)
+
+        cube(frame, landmarks.part(36), landmarks.part(45), a1, a2, a3)
+    cv2.imshow("Frame", frame)
+
+    key=cv2.waitKey(1)&0xFF
+    if key==ord('q'):
+        break
+
+cap.release()
+cv2.destroyAllWindows()
+1

Divers/tutoriel20/orientation.py

@@ -0,0 +1,66 @@
+import cv2
+import numpy as np
+import dlib
+import math
+
+cap=cv2.VideoCapture(0)
+
+detector=dlib.get_frontal_face_detector()
+predictor=dlib.shape_predictor("shape_predictor_68_face_landmarks.dat")
+
+while True:
+    ret, frame=cap.read()
+    gray=cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
+
+    faces=detector(gray)
+    if faces is not None:
+        i=np.zeros(shape=(frame.shape), dtype=np.uint8)
+    for face in faces:
+        landmarks=predictor(gray, face)
+
+        d_eyes=math.sqrt(math.pow(landmarks.part(36).x-landmarks.part(45).x, 2)+math.pow(landmarks.part(36).y-landmarks.part(45).y, 2))
+        d1=math.sqrt(math.pow(landmarks.part(36).x-landmarks.part(30).x, 2)+math.pow(landmarks.part(36).y-landmarks.part(30).y, 2))
+        d2=math.sqrt(math.pow(landmarks.part(45).x-landmarks.part(30).x, 2)+math.pow(landmarks.part(45).y-landmarks.part(30).y, 2))
+        coeff=d1+d2
+
+        a1=int(250*(landmarks.part(36).y-landmarks.part(45).y)/coeff)
+        a2=int(250*(d1-d2)/coeff)
+        cosb=min((math.pow(d2, 2)-math.pow(d1, 2)+math.pow(d_eyes, 2))/(2*d2*d_eyes), 1)
+        a3=int(250*(d2*math.sin(math.acos(cosb))-coeff/4)/coeff)
+
+        for n in range(0, 68):
+            x=landmarks.part(n).x
+            y=landmarks.part(n).y
+            if n==30 or n==36 or n==45:
+                cv2.circle(i, (x, y), 3, (255, 255, 0), -1)
+            else:
+                cv2.circle(i, (x, y), 3, (255, 0, 0), -1)
+        print("{:+05d} {:+05d} {:+05d}".format(a1, a2, a3))
+        flag=1
+        txt="Laurent regarde "
+        if a2<-40:
+            txt+="a droite "
+            flag=0
+        if a2>40:
+            txt+="a gauche "
+            flag=0
+        if a3<-10:
+            txt+="en haut "
+            flag=0
+        if a3>10:
+            txt+="en bas "
+            flag=0
+        if flag:
+            txt+="la camera "
+        if a1<-40:
+            txt+="et incline la tete a gauche "
+        if a1>40:
+            txt+="et incline la tete a droite "
+        cv2.putText(frame, txt, (10, 30), cv2.FONT_HERSHEY_PLAIN, 1.2, (255, 255, 255), 2)
+    cv2.imshow("Frame", frame)
+    key=cv2.waitKey(1)&0xFF
+    if key==ord('q'):
+        break
+
+cap.release()
+cv2.destroyAllWindows()

Divers/tutoriel20/visage_point.py

@@ -0,0 +1,35 @@
+import cv2
+import numpy as np
+import dlib
+import math
+
+cap=cv2.VideoCapture(0)
+detector=dlib.get_frontal_face_detector()
+predictor=dlib.shape_predictor("shape_predictor_68_face_landmarks.dat")
+
+while True:
+    ret, frame=cap.read()
+    tickmark=cv2.getTickCount()
+    gray=cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
+    faces=detector(gray)
+    for face in faces:
+        landmarks=predictor(gray, face)
+        i=np.zeros(shape=(frame.shape), dtype=np.uint8)
+        for n in range(0, 68):
+            x=landmarks.part(n).x
+            y=landmarks.part(n).y
+            cv2.circle(frame, (x, y), 3, (255, 0, 0), -1)
+            if n==30 or n==36 or n==45:
+                cv2.circle(i, (x, y), 3, (255, 255, 0), -1)
+            else:
+                cv2.circle(i, (x, y), 3, (255, 0, 0), -1)
+        cv2.imshow("i", i)
+    fps=cv2.getTickFrequency()/(cv2.getTickCount()-tickmark)
+    cv2.putText(frame, "FPS: {:05.2f}".format(fps), (10, 30), cv2.FONT_HERSHEY_PLAIN, 2, (255, 0, 0), 2)
+    cv2.imshow("Frame", frame)
+    key=cv2.waitKey(1)&0xFF
+    if key==ord('q'):
+        break
+
+cap.release()
+cv2.destroyAllWindows()

Divers/tutoriel20/visage_rectangle.py

@@ -0,0 +1,27 @@
+import cv2
+import numpy as np
+import dlib
+
+cap=cv2.VideoCapture(0)
+detector=dlib.get_frontal_face_detector()
+
+while True:
+    ret, frame=cap.read()
+    tickmark=cv2.getTickCount()
+    gray=cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
+    faces=detector(gray)
+    for face in faces:
+        x1=face.left()
+        y1=face.top()
+        x2=face.right()
+        y2=face.bottom()
+        cv2.rectangle(frame, (x1, y1), (x2, y2), (255, 0, 0), 2)
+    fps=cv2.getTickFrequency()/(cv2.getTickCount()-tickmark)
+    cv2.putText(frame, "FPS: {:05.2f}".format(fps), (10, 30), cv2.FONT_HERSHEY_PLAIN, 2, (255, 0, 0), 2)
+    cv2.imshow("Frame", frame)
+    key=cv2.waitKey(1)&0xFF
+    if key==ord('q'):
+        break
+
+cap.release()
+cv2.destroyAllWindows()

Divers/tutoriel25-2/README.md

@@ -0,0 +1,8 @@
+# Tutoriel 25
+## Lecture des panneaux de limitation de vitesse
+
+Les vidéos de ce tutoriel sont disponibles aux adresses suivantes:<br>
+partie 1: https://www.youtube.com/watch?v=PvD5POjXw8Q <br>
+partie 2: https://www.youtube.com/watch?v=TYDi0SNCUr0 <br>
+partie 3: https://www.youtube.com/watch?v=fBysd-Y17Tw
+

Divers/tutoriel25-2/common.py

@@ -0,0 +1,89 @@
+import tensorflow as tf
+from tensorflow.keras import layers, models
+import os
+import cv2
+
+size=42
+dir_images_panneaux="images_panneaux"
+dir_images_autres_panneaux="images_autres_panneaux"
+dir_images_sans_panneaux="images_sans_panneaux"
+
+def panneau_model(nbr_classes):
+    model=tf.keras.Sequential()
+
+    model.add(layers.Input(shape=(size, size, 3), dtype='float32'))
+    
+    model.add(layers.Conv2D(128, 3, strides=1))
+    model.add(layers.Dropout(0.2))
+    model.add(layers.BatchNormalization())
+    model.add(layers.Activation('relu'))
+
+    model.add(layers.Conv2D(128, 3, strides=1))
+    model.add(layers.Dropout(0.2))
+    model.add(layers.BatchNormalization())
+    model.add(layers.Activation('relu'))
+
+    model.add(layers.MaxPool2D(pool_size=2, strides=2))
+
+    model.add(layers.Conv2D(256, 3, strides=1))
+    model.add(layers.Dropout(0.3))
+    model.add(layers.BatchNormalization())
+    model.add(layers.Activation('relu'))
+
+    model.add(layers.Conv2D(256, 3, strides=1))
+    model.add(layers.Dropout(0.4))
+    model.add(layers.BatchNormalization())
+    model.add(layers.Activation('relu'))
+
+    model.add(layers.MaxPool2D(pool_size=2, strides=2))
+
+    model.add(layers.Flatten())
+    model.add(layers.Dense(512, activation='relu'))
+    model.add(layers.Dropout(0.5))
+    model.add(layers.BatchNormalization())
+    model.add(layers.Dense(nbr_classes, activation='softmax'))
+
+    return model
+
+def is_panneau_model():
+    model=tf.keras.Sequential()
+
+    model.add(layers.Input(shape=(size, size, 3), dtype='float32'))
+    
+    model.add(layers.Conv2D(64, 3, strides=1))
+    model.add(layers.BatchNormalization())
+    model.add(layers.Activation('relu'))
+    model.add(layers.MaxPool2D(pool_size=2, strides=2))
+
+    model.add(layers.Conv2D(128, 3, strides=1))
+    model.add(layers.BatchNormalization())
+    model.add(layers.Activation('relu'))
+    model.add(layers.MaxPool2D(pool_size=2, strides=2))
+
+    model.add(layers.Flatten())
+    model.add(layers.Dense(256, activation='relu'))
+    model.add(layers.BatchNormalization())
+    model.add(layers.Dense(1, activation='sigmoid'))
+
+    return model
+
+def lire_images_panneaux(dir_images_panneaux, size=None):
+    tab_panneau=[]
+    tab_image_panneau=[]
+
+    if not os.path.exists(dir_images_panneaux):
+        quit("Le repertoire d'image n'existe pas: {}".format(dir_images_panneaux))
+
+    files=os.listdir(dir_images_panneaux)
+    if files is None:
+        quit("Le repertoire d'image est vide: {}".format(dir_images_panneaux))
+
+    for file in sorted(files):
+        if file.endswith("png"):
+            tab_panneau.append(file.split(".")[0])
+            image=cv2.imread(dir_images_panneaux+"/"+file)
+            if size is not None:
+                image=cv2.resize(image, (size, size), cv2.INTER_LANCZOS4)
+            tab_image_panneau.append(image)
+            
+    return tab_panneau, tab_image_panneau

Divers/tutoriel25-2/dataset.py

@@ -0,0 +1,82 @@
+import numpy as np
+import cv2
+from multiprocessing import Pool
+import multiprocessing
+import random
+
+def bruit(image_orig):
+    h, w, c=image_orig.shape
+    n=np.random.randn(h, w, c)*random.randint(5, 30)
+    return np.clip(image_orig+n, 0, 255).astype(np.uint8)
+
+def change_gamma(image, alpha=1.0, beta=0.0):
+    return np.clip(alpha*image+beta, 0, 255).astype(np.uint8)
+    
+def modif_img(img):
+    h, w, c=img.shape
+
+    r_color=[np.random.randint(255), np.random.randint(255), np.random.randint(255)]
+    img=np.where(img==[142, 142, 142], r_color, img).astype(np.uint8)
+
+    if np.random.randint(3):
+        k_max=3
+        kernel_blur=np.random.randint(k_max)*2+1
+        img=cv2.GaussianBlur(img, (kernel_blur, kernel_blur), 0)
+
+    M=cv2.getRotationMatrix2D((int(w/2), int(h/2)), random.randint(-10, 10), 1)
+    img=cv2.warpAffine(img, M, (w, h))
+        
+    if np.random.randint(2):
+        a=int(max(w, h)/5)+1
+        pts1=np.float32([[0, 0], [w, 0], [0, h], [w, h]])
+        pts2=np.float32([[0+random.randint(-a, a), 0+random.randint(-a, a)], [w-random.randint(-a, a), 0+random.randint(-a, a)], [0+random.randint(-a, a), h-random.randint(-a, a)], [w-random.randint(-a, a), h-random.randint(-a, a)]])        
+        M=cv2.getPerspectiveTransform(pts1,pts2)
+        img=cv2.warpPerspective(img, M, (w, h))
+        
+    if np.random.randint(2):
+        r=random.randint(0, 5)
+        h2=int(h*0.9)
+        w2=int(w*0.9)
+        if r==0:
+            img=img[0:w2, 0:h2]
+        elif r==1:
+            img=img[w-w2:w, 0:h2]
+        elif r==2:
+            img=img[0:w2, h-h2:h]
+        elif r==3:
+            img=img[w-w2:w, h-h2:h]
+        img=cv2.resize(img, (h, w))
+
+    if np.random.randint(2):
+        r=random.randint(1, int(max(w, h)*0.15))
+        img=img[r:w-r, r:h-r]
+        img=cv2.resize(img, (h, w))
+
+    if not np.random.randint(4):
+        t=np.empty((h, w, c) , dtype=np.float32)
+        for i in range(h):
+            for j in range(w):
+                for k in range(c):
+                    t[i][j][k]=(i/h)
+        M=cv2.getRotationMatrix2D((int(w/2), int(h/2)), np.random.randint(4)*90, 1)
+        t=cv2.warpAffine(t, M, (w, h))
+        img=(cv2.multiply((img/255).astype(np.float32), t)*255).astype(np.uint8)
+
+    img=change_gamma(img, random.uniform(0.6, 1.0), -np.random.randint(50))
+
+    if not np.random.randint(4):
+        p=(15+np.random.randint(10))/100
+        img=(img*p+50*(1-p)).astype(np.uint8)+np.random.randint(100)
+
+    img=bruit(img)
+        
+    return img
+
+def create_lot_img(image, nbr, nbr_thread=None):
+    if nbr_thread is None:
+        nbr_thread=multiprocessing.cpu_count()
+    lot_original=np.repeat([image], nbr, axis=0)
+    with Pool(nbr_thread) as p:
+        lot_result=p.map(modif_img, lot_original)
+        p.close()
+    return lot_result

Divers/tutoriel25-2/genere_fond.py

@@ -0,0 +1,40 @@
+import cv2
+import numpy as np
+import random
+import os
+import common
+
+video="videos/France_Motorway.mp4"
+
+if not os.path.isdir(common.dir_images_sans_panneaux):
+    os.mkdir(common.dir_images_sans_panneaux)
+
+if not os.path.exists(video):
+    print("Vidéo non présente:", video)
+    quit()
+    
+cap=cv2.VideoCapture(video)
+
+id=0
+nbr_image=100000
+
+nbr_image_par_frame=int(100000/cap.get(cv2.CAP_PROP_FRAME_COUNT))+1
+
+while True:
+    ret, frame=cap.read()
+    if ret is False:
+        quit()
+    h, w, c=frame.shape
+
+    for cpt in range(nbr_image_par_frame):
+        x=random.randint(0, w-common.size)
+        y=random.randint(0, h-common.size)
+        img=frame[y:y+common.size, x:x+common.size]        
+        cv2.imwrite(common.dir_images_sans_panneaux+"/{:d}.png".format(id), img)
+        id+=1
+        if id==nbr_image:
+            quit()
+    
+
+
+        

Divers/tutoriel25-2/genere_panneaux.py

@@ -0,0 +1,29 @@
+import numpy as np
+from sklearn.utils import shuffle
+import cv2
+import common
+import dataset
+
+tab_panneau, tab_image_panneau=common.lire_images_panneaux(common.dir_images_panneaux, common.size)
+
+tab_images=np.array([]).reshape(0, common.size, common.size, 3)
+tab_labels=[]
+
+id=0
+for image in tab_image_panneau:
+    lot=dataset.create_lot_img(image, 1000)
+    tab_images=np.concatenate([tab_images, lot])
+    tab_labels=np.concatenate([tab_labels, np.full(len(lot), id)])
+    id+=1
+
+tab_panneau=np.array(tab_panneau)
+tab_images=np.array(tab_images, dtype=np.float32)/255
+tab_labels=np.array(tab_labels).reshape([-1, 1])
+
+tab_images, tab_labels=shuffle(tab_images, tab_labels)
+
+for i in range(len(tab_images)):
+    cv2.imshow("panneau", tab_images[i])
+    print("label", tab_labels[i], "panneau", tab_panneau[int(tab_labels[i])])
+    if cv2.waitKey()&0xFF==ord('q'):
+        quit()

Divers/tutoriel25-2/houghcircles.py

@@ -0,0 +1,36 @@
+import cv2
+import numpy as np
+
+param1=30
+param2=55
+dp=1.0
+
+cap=cv2.VideoCapture(0)
+
+while True:
+    ret, frame=cap.read()
+    gray=cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
+    circles=cv2.HoughCircles(gray, cv2.HOUGH_GRADIENT, dp, 20, param1=param1, param2=param2, minRadius=10, maxRadius=50)
+    if circles is not None:
+        circles=np.around(circles).astype(np.int32)
+        for i in circles[0, :]:
+            if i[2]!=0:
+                cv2.circle(frame, (i[0], i[1]), i[2], (0, 255, 0), 4)
+    cv2.putText(frame, "[i|k]dp: {:4.2f}  [o|l]param1: {:d}  [p|m]param2: {:d}".format(dp, param1, param2), (10, 40), cv2.FONT_HERSHEY_COMPLEX_SMALL, 1, (255, 0, 0), 1)
+    cv2.imshow("Video", frame)
+    key=cv2.waitKey(1)&0xFF
+    if key==ord('q'):
+        quit()
+    if key==ord('i'):
+        dp=min(10, dp+0.1)
+    if key==ord('k'):
+        dp=max(0.1, dp-0.1)
+    if key==ord('o'):
+        param1=min(255, param1+1)
+    if key==ord('l'):
+        param1=max(1, param1-1)
+    if key==ord('p'):
+        param2=min(255, param2+1)
+    if key==ord('m'):
+        param2=max(1, param2-1)
+cv2.destroyAllWindows()

Divers/tutoriel25-2/lire_panneau.py

@@ -0,0 +1,70 @@
+import tensorflow as tf
+import cv2
+import os
+import numpy as np
+import random
+import common
+
+th1=30
+th2=55
+
+video_dir="dashcam Cedric"
+
+tab_panneau, tab_image_panneau=common.lire_images_panneaux(common.dir_images_panneaux)
+
+model_is_panneau=common.is_panneau_model()
+checkpoint=tf.train.Checkpoint(model_is_panneau=model_is_panneau)
+checkpoint.restore(tf.train.latest_checkpoint("./training_is_panneau/"))
+
+model_panneau=common.panneau_model(len(tab_panneau))
+checkpoint=tf.train.Checkpoint(model_panneau=model_panneau)
+checkpoint.restore(tf.train.latest_checkpoint("./training_panneau/"))
+
+l=os.listdir(video_dir)
+random.shuffle(l)
+
+for video in l:
+    if not video.endswith("mp4"):
+        continue
+    cap=cv2.VideoCapture(video_dir+"/"+video)
+
+    print("video:", video)
+    id_panneau=-1    
+    while True:
+        ret, frame=cap.read()
+        if ret is False:
+            break
+        f_w, f_h, f_c=frame.shape
+        frame=cv2.resize(frame, (int(f_h/1.5), int(f_w/1.5)))
+
+        image=frame[200:400, 700:1000]
+        cv2.rectangle(frame, (700, 200), (1000, 400), (255, 255, 255), 1)
+        gray=cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
+        circles=cv2.HoughCircles(gray, cv2.HOUGH_GRADIENT, 1, 20, param1=th1, param2=th2, minRadius=5, maxRadius=45)
+        if circles is not None:
+            circles=np.int16(np.around(circles))
+            for i in circles[0,:]:
+                if i[2]!=0:
+                    panneau=cv2.resize(image[max(0, i[1]-i[2]):i[1]+i[2], max(0, i[0]-i[2]):i[0]+i[2]], (common.size, common.size))/255
+                    cv2.imshow("panneau", panneau)
+                    prediction=model_is_panneau(np.array([panneau]), training=False)
+                    print("prediction", prediction)
+                    if prediction[0][0]>0.9:
+                        prediction=model_panneau(np.array([panneau]), training=False)
+                        id_panneau=np.argmax(prediction[0])
+                        print("panneau", prediction, id_panneau, tab_panneau[id_panneau])
+                        w, h, c=tab_image_panneau[id_panneau].shape
+        if id_panneau!=-1:
+            frame[0:h, 0:w, :]=tab_image_panneau[id_panneau]
+        cv2.putText(frame, "fichier:"+video, (30, 30), cv2.FONT_HERSHEY_DUPLEX, 1, (0, 255, 0), 1, cv2.LINE_AA)
+        cv2.imshow("Video", frame)
+        key=cv2.waitKey(1)&0xFF
+        if key==ord('q'):
+            quit()
+        if key==ord('a'):
+            for cpt in range(100):
+                ret, frame=cap.read()
+        if key==ord('f'):
+            break
+                
+cv2.destroyAllWindows()

Divers/tutoriel25-2/train_is_panneau.py

@@ -0,0 +1,135 @@
+import tensorflow as tf
+from tensorflow.keras import layers, models
+import numpy as np
+from sklearn.utils import shuffle
+from sklearn.model_selection import train_test_split
+import cv2
+import os
+import common
+import time
+import dataset
+
+batch_size=64
+nbr_entrainement=20
+
+tab_images=np.array([]).reshape(0, common.size, common.size, 3)
+
+tab_panneau, tab_image_panneau=common.lire_images_panneaux(common.dir_images_panneaux, common.size)
+
+if not os.path.exists(common.dir_images_autres_panneaux):
+    quit("Le repertoire d'image n'existe pas: {}".format(common.dir_images_autres_panneaux))
+
+if not os.path.exists(common.dir_images_sans_panneaux):
+    quit("Le repertoire d'image n'existe pas:".format(common.dir_images_sans_panneaux))
+
+nbr=0
+for image in tab_image_panneau:
+    lot=dataset.create_lot_img(image, 12000)
+    tab_images=np.concatenate([tab_images, lot])
+    nbr+=len(lot)
+
+tab_labels=np.full(nbr, 1)
+
+print("Image panneaux:", nbr)
+
+files=os.listdir(common.dir_images_autres_panneaux)
+if files is None:
+    quit("Le repertoire d'image est vide:".format(common.dir_images_autres_panneaux))
+
+nbr=0
+for file in files:
+    if file.endswith("png"):
+        path=os.path.join(common.dir_images_autres_panneaux, file)
+        image=cv2.resize(cv2.imread(path), (common.size, common.size), cv2.INTER_LANCZOS4)
+        lot=dataset.create_lot_img(image, 700)
+        tab_images=np.concatenate([tab_images, lot])
+        nbr+=len(lot)
+
+tab_labels=np.concatenate([tab_labels, np.full(nbr, 0)])
+        
+print("Image autres panneaux:", nbr)
+
+nbr_np=int(len(tab_images)/2)
+print("nbr_np", nbr_np)
+
+id=1
+nbr=0
+tab=[]
+for cpt in range(nbr_np):
+    file=common.dir_images_sans_panneaux+"/{:d}.png".format(id)
+    if not os.path.isfile(file):
+        break
+    image=cv2.resize(cv2.imread(file), (common.size, common.size))
+    tab.append(image)
+    id+=1
+    nbr+=1
+
+tab_images=np.concatenate([tab_images, tab])
+tab_labels=np.concatenate([tab_labels, np.full(nbr, 0)])
+print("Image sans panneaux:", nbr)
+
+tab_images=np.array(tab_images, dtype=np.float32)/255
+tab_labels=np.array(tab_labels, dtype=np.float32).reshape([-1, 1])
+
+tab_images, tab_labels=shuffle(tab_images, tab_labels)
+train_images, test_images, train_labels, test_labels=train_test_split(tab_images, tab_labels, test_size=0.10)
+
+train_ds=tf.data.Dataset.from_tensor_slices((train_images, train_labels)).batch(batch_size)
+test_ds=tf.data.Dataset.from_tensor_slices((test_images, test_labels)).batch(batch_size)
+
+print("train_images", len(train_images))
+print("test_images", len(test_images))
+print("nbr panneau", len(np.where(train_labels==0.)[1]), train_labels.shape)
+
+@tf.function
+def train_step(images, labels):
+  with tf.GradientTape() as tape:
+    predictions=model_is_panneau(images)
+    loss=loss_object(labels, predictions)
+  gradients=tape.gradient(loss, model_is_panneau.trainable_variables)
+  optimizer.apply_gradients(zip(gradients, model_is_panneau.trainable_variables))
+  train_loss(loss)
+  train_accuracy(labels, predictions)
+
+def train(train_ds, nbr_entrainement):
+  for entrainement in range(nbr_entrainement):
+    start=time.time()
+    for images, labels in train_ds:
+      train_step(images, labels)
+    message='Entrainement {:04d}, loss: {:6.4f}, accuracy: {:7.4f}%, temps: {:7.4f}'
+    print(message.format(entrainement+1,
+                         train_loss.result(),
+                         train_accuracy.result()*100,
+                         time.time()-start))
+    train_loss.reset_states()
+    train_accuracy.reset_states()
+    test(test_ds)
+
+def test(test_ds):
+  start=time.time()
+  for test_images, test_labels in test_ds:
+    predictions=model_is_panneau(test_images)
+    t_loss=loss_object(test_labels, predictions)
+    test_loss(t_loss)
+    test_accuracy(test_labels, predictions)
+  message='   >>> Test: loss: {:6.4f}, accuracy: {:7.4f}%, temps: {:7.4f}'
+  print(message.format(test_loss.result(),
+                       test_accuracy.result()*100,
+                       time.time()-start))
+  test_loss.reset_states()
+  test_accuracy.reset_states()
+  
+optimizer=tf.keras.optimizers.Adam()
+loss_object=tf.keras.losses.BinaryCrossentropy()
+train_loss=tf.keras.metrics.Mean()
+train_accuracy=tf.keras.metrics.BinaryAccuracy()
+test_loss=tf.keras.metrics.Mean()
+test_accuracy=tf.keras.metrics.BinaryAccuracy()
+model_is_panneau=common.is_panneau_model()
+checkpoint=tf.train.Checkpoint(model_is_panneau=model_is_panneau)
+
+print("Entrainement")
+train(train_ds, nbr_entrainement)
+test(test_ds)
+
+checkpoint.save(file_prefix="./training_is_panneau/is_panneau")

Divers/tutoriel25-2/train_panneau.py

@@ -0,0 +1,95 @@
+import tensorflow as tf
+from tensorflow.keras import layers, models
+import numpy as np
+from sklearn.utils import shuffle
+from sklearn.model_selection import train_test_split
+import cv2
+import os
+import time
+import common
+import dataset
+
+batch_size=128
+nbr_entrainement=20
+
+tab_images=np.array([]).reshape(0, common.size, common.size, 3)
+tab_labels=[]
+
+tab_panneau, tab_image_panneau=common.lire_images_panneaux(common.dir_images_panneaux, common.size)
+
+id=0
+for image in tab_image_panneau:
+    lot=dataset.create_lot_img(image, 12000)
+    tab_images=np.concatenate((tab_images, lot))
+    tab_labels=np.concatenate([tab_labels, np.full(len(lot), id)])
+    id+=1
+                
+tab_panneau=np.array(tab_panneau)
+tab_images=np.array(tab_images, dtype=np.float32)/255
+tab_labels=np.array(tab_labels, dtype=np.float32).reshape([-1, 1])
+
+train_images, test_images, train_labels, test_labels=train_test_split(tab_images, tab_labels, test_size=0.10)
+
+train_ds=tf.data.Dataset.from_tensor_slices((train_images, train_labels)).batch(batch_size)
+test_ds=tf.data.Dataset.from_tensor_slices((test_images, test_labels)).batch(batch_size)
+
+print("train_images", len(train_images))
+print("test_images", len(test_images))
+
+@tf.function
+def train_step(images, labels):
+  with tf.GradientTape() as tape:
+    predictions=model_panneau(images)
+    loss=loss_object(labels, predictions)
+  gradients=tape.gradient(loss, model_panneau.trainable_variables)
+  optimizer.apply_gradients(zip(gradients, model_panneau.trainable_variables))
+  train_loss(loss)
+  train_accuracy(labels, predictions)
+
+def train(train_ds, nbr_entrainement):
+  for entrainement in range(nbr_entrainement):
+    start=time.time()
+    for images, labels in train_ds:
+      train_step(images, labels)
+    message='Entrainement {:04d}: loss: {:6.4f}, accuracy: {:7.4f}%, temps: {:7.4f}'
+    print(message.format(entrainement+1,
+                         train_loss.result(),
+                         train_accuracy.result()*100,
+                         time.time()-start))
+    train_loss.reset_states()
+    train_accuracy.reset_states()
+    test(test_ds)
+    
+def test(test_ds):
+  start=time.time()
+  for test_images, test_labels in test_ds:
+    predictions=model_panneau(test_images)
+    t_loss=loss_object(test_labels, predictions)
+    test_loss(t_loss)
+    test_accuracy(test_labels, predictions)
+  message='   >>> Test: loss: {:6.4f}, accuracy: {:7.4f}%, temps: {:7.4f}'
+  print(message.format(test_loss.result(),
+                       test_accuracy.result()*100,
+                       time.time()-start))
+  test_loss.reset_states()
+  test_accuracy.reset_states()
+
+optimizer=tf.keras.optimizers.Adam()
+loss_object=tf.keras.losses.SparseCategoricalCrossentropy()
+train_loss=tf.keras.metrics.Mean()
+train_accuracy=tf.keras.metrics.SparseCategoricalAccuracy()
+test_loss=tf.keras.metrics.Mean()
+test_accuracy=tf.keras.metrics.SparseCategoricalAccuracy()
+model_panneau=common.panneau_model(len(tab_panneau))
+checkpoint=tf.train.Checkpoint(model_panneau=model_panneau)
+
+print("Entrainement")
+train(train_ds, nbr_entrainement)
+checkpoint.save(file_prefix="./training_panneau/panneau")
+
+for i in range(len(test_images)):
+    prediction=model_panneau(np.array([test_images[i]]))
+    print("prediction", prediction, tab_panneau[np.argmax(prediction[0])])
+    cv2.imshow("image", test_images[i])
+    if cv2.waitKey()&0xFF==ord('q'):
+        break

Divers/tutoriel25-3/README.md

@@ -0,0 +1,8 @@
+# Tutoriel 25
+## Lecture des panneaux de limitation de vitesse
+
+Les vidéos de ce tutoriel sont disponibles aux adresses suivantes:<br>
+partie 1: https://www.youtube.com/watch?v=PvD5POjXw8Q <br>
+partie 2: https://www.youtube.com/watch?v=TYDi0SNCUr0 <br>
+partie 3: https://www.youtube.com/watch?v=fBysd-Y17Tw
+

Divers/tutoriel25-3/common.py

@@ -0,0 +1,67 @@
+import tensorflow as tf
+from tensorflow.keras import layers, models
+import os
+import cv2
+
+size=42
+dir_images_panneaux="images_panneaux"
+dir_images_autres_panneaux="images_autres_panneaux"
+dir_images_sans_panneaux="images_sans_panneaux"
+
+def panneau_model(nbr_classes):
+    model=tf.keras.Sequential()
+
+    model.add(layers.Input(shape=(size, size, 3), dtype='float32'))
+    
+    model.add(layers.Conv2D(128, 3, strides=1))
+    model.add(layers.Dropout(0.2))
+    model.add(layers.BatchNormalization())
+    model.add(layers.Activation('relu'))
+
+    model.add(layers.Conv2D(128, 3, strides=1))
+    model.add(layers.Dropout(0.2))
+    model.add(layers.BatchNormalization())
+    model.add(layers.Activation('relu'))
+
+    model.add(layers.MaxPool2D(pool_size=2, strides=2))
+
+    model.add(layers.Conv2D(256, 3, strides=1))
+    model.add(layers.Dropout(0.3))
+    model.add(layers.BatchNormalization())
+    model.add(layers.Activation('relu'))
+
+    model.add(layers.Conv2D(256, 3, strides=1))
+    model.add(layers.Dropout(0.4))
+    model.add(layers.BatchNormalization())
+    model.add(layers.Activation('relu'))
+
+    model.add(layers.MaxPool2D(pool_size=2, strides=2))
+
+    model.add(layers.Flatten())
+    model.add(layers.Dense(512, activation='relu'))
+    model.add(layers.Dropout(0.5))
+    model.add(layers.BatchNormalization())
+    model.add(layers.Dense(nbr_classes, activation='sigmoid'))
+    
+    return model
+
+def lire_images_panneaux(dir_images_panneaux, size=None):
+    tab_panneau=[]
+    tab_image_panneau=[]
+
+    if not os.path.exists(dir_images_panneaux):
+        quit("Le repertoire d'image n'existe pas: {}".format(dir_images_panneaux))
+
+    files=os.listdir(dir_images_panneaux)
+    if files is None:
+        quit("Le repertoire d'image est vide: {}".format(dir_images_panneaux))
+
+    for file in sorted(files):
+        if file.endswith("png"):
+            tab_panneau.append(file.split(".")[0])
+            image=cv2.imread(dir_images_panneaux+"/"+file)
+            if size is not None:
+                image=cv2.resize(image, (size, size), cv2.INTER_LANCZOS4)
+            tab_image_panneau.append(image)
+            
+    return tab_panneau, tab_image_panneau

Divers/tutoriel25-3/dataset.py

@@ -0,0 +1,82 @@
+import numpy as np
+import cv2
+from multiprocessing import Pool
+import multiprocessing
+import random
+
+def bruit(image_orig):
+    h, w, c=image_orig.shape
+    n=np.random.randn(h, w, c)*random.randint(5, 30)
+    return np.clip(image_orig+n, 0, 255).astype(np.uint8)
+
+def change_gamma(image, alpha=1.0, beta=0.0):
+    return np.clip(alpha*image+beta, 0, 255).astype(np.uint8)
+    
+def modif_img(img):
+    h, w, c=img.shape
+
+    r_color=[np.random.randint(255), np.random.randint(255), np.random.randint(255)]
+    img=np.where(img==[142, 142, 142], r_color, img).astype(np.uint8)
+
+    if np.random.randint(3):
+        k_max=3
+        kernel_blur=np.random.randint(k_max)*2+1
+        img=cv2.GaussianBlur(img, (kernel_blur, kernel_blur), 0)
+
+    M=cv2.getRotationMatrix2D((int(w/2), int(h/2)), random.randint(-10, 10), 1)
+    img=cv2.warpAffine(img, M, (w, h))
+        
+    if np.random.randint(2):
+        a=int(max(w, h)/5)+1
+        pts1=np.float32([[0, 0], [w, 0], [0, h], [w, h]])
+        pts2=np.float32([[0+random.randint(-a, a), 0+random.randint(-a, a)], [w-random.randint(-a, a), 0+random.randint(-a, a)], [0+random.randint(-a, a), h-random.randint(-a, a)], [w-random.randint(-a, a), h-random.randint(-a, a)]])        
+        M=cv2.getPerspectiveTransform(pts1,pts2)
+        img=cv2.warpPerspective(img, M, (w, h))
+        
+    if np.random.randint(2):
+        r=random.randint(0, 5)
+        h2=int(h*0.9)
+        w2=int(w*0.9)
+        if r==0:
+            img=img[0:w2, 0:h2]
+        elif r==1:
+            img=img[w-w2:w, 0:h2]
+        elif r==2:
+            img=img[0:w2, h-h2:h]
+        elif r==3:
+            img=img[w-w2:w, h-h2:h]
+        img=cv2.resize(img, (h, w))
+
+    if np.random.randint(2):
+        r=random.randint(1, int(max(w, h)*0.15))
+        img=img[r:w-r, r:h-r]
+        img=cv2.resize(img, (h, w))
+
+    if not np.random.randint(4):
+        t=np.empty((h, w, c) , dtype=np.float32)
+        for i in range(h):
+            for j in range(w):
+                for k in range(c):
+                    t[i][j][k]=(i/h)
+        M=cv2.getRotationMatrix2D((int(w/2), int(h/2)), np.random.randint(4)*90, 1)
+        t=cv2.warpAffine(t, M, (w, h))
+        img=(cv2.multiply((img/255).astype(np.float32), t)*255).astype(np.uint8)
+
+    img=change_gamma(img, random.uniform(0.6, 1.0), -np.random.randint(50))
+
+    if not np.random.randint(4):
+        p=(15+np.random.randint(10))/100
+        img=(img*p+50*(1-p)).astype(np.uint8)+np.random.randint(100)
+
+    img=bruit(img)
+        
+    return img
+
+def create_lot_img(image, nbr, nbr_thread=None):
+    if nbr_thread is None:
+        nbr_thread=multiprocessing.cpu_count()
+    lot_original=np.repeat([image], nbr, axis=0)
+    with Pool(nbr_thread) as p:
+        lot_result=p.map(modif_img, lot_original)
+        p.close()
+    return lot_result

Divers/tutoriel25-3/genere_fond.py

@@ -0,0 +1,40 @@
+import cv2
+import numpy as np
+import random
+import os
+import common
+
+video="videos/France_Motorway.mp4"
+
+if not os.path.isdir(common.dir_images_sans_panneaux):
+    os.mkdir(common.dir_images_sans_panneaux)
+
+if not os.path.exists(video):
+    print("Vidéo non présente:", video)
+    quit()
+    
+cap=cv2.VideoCapture(video)
+
+id=0
+nbr_image=100000
+
+nbr_image_par_frame=int(100000/cap.get(cv2.CAP_PROP_FRAME_COUNT))+1
+
+while True:
+    ret, frame=cap.read()
+    if ret is False:
+        quit()
+    h, w, c=frame.shape
+
+    for cpt in range(nbr_image_par_frame):
+        x=random.randint(0, w-common.size)
+        y=random.randint(0, h-common.size)
+        img=frame[y:y+common.size, x:x+common.size]        
+        cv2.imwrite(common.dir_images_sans_panneaux+"/{:d}.png".format(id), img)
+        id+=1
+        if id==nbr_image:
+            quit()
+    
+
+
+        

Divers/tutoriel25-3/genere_panneaux.py

@@ -0,0 +1,29 @@
+import numpy as np
+from sklearn.utils import shuffle
+import cv2
+import common
+import dataset
+
+tab_panneau, tab_image_panneau=common.lire_images_panneaux(common.dir_images_panneaux, common.size)
+
+tab_images=np.array([]).reshape(0, common.size, common.size, 3)
+tab_labels=[]
+
+id=0
+for image in tab_image_panneau:
+    lot=dataset.create_lot_img(image, 1000)
+    tab_images=np.concatenate([tab_images, lot])
+    tab_labels=np.concatenate([tab_labels, np.full(len(lot), id)])
+    id+=1
+
+tab_panneau=np.array(tab_panneau)
+tab_images=np.array(tab_images, dtype=np.float32)/255
+tab_labels=np.array(tab_labels).reshape([-1, 1])
+
+tab_images, tab_labels=shuffle(tab_images, tab_labels)
+
+for i in range(len(tab_images)):
+    cv2.imshow("panneau", tab_images[i])
+    print("label", tab_labels[i], "panneau", tab_panneau[int(tab_labels[i])])
+    if cv2.waitKey()&0xFF==ord('q'):
+        quit()

Divers/tutoriel25-3/lire_panneau.py

@@ -0,0 +1,69 @@
+import tensorflow as tf
+import cv2
+import os
+import numpy as np
+import random
+import common
+
+th1=30
+th2=55
+
+video_dir="d:/dashcam Cedric"
+
+tab_panneau, tab_image_panneau=common.lire_images_panneaux(common.dir_images_panneaux)
+
+model_panneau=common.panneau_model(len(tab_panneau))
+checkpoint=tf.train.Checkpoint(model_panneau=model_panneau)
+checkpoint.restore(tf.train.latest_checkpoint("./training_panneau/"))
+
+l=os.listdir(video_dir)
+random.shuffle(l)
+
+for video in l:
+    if not video.endswith("mp4"):
+        continue
+    cap=cv2.VideoCapture(video_dir+"/"+video)
+
+    print("video:", video)
+    id_panneau=-1
+    while True:
+        ret, frame=cap.read()
+        if ret is False:
+            break
+        f_w, f_h, f_c=frame.shape
+        frame=cv2.resize(frame, (int(f_h/1.5), int(f_w/1.5)))
+
+        image=frame[200:400, 700:1000]
+        cv2.rectangle(frame, (700, 200), (1000, 400), (255, 255, 255), 1)
+
+        gray=cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
+
+        circles=cv2.HoughCircles(gray, cv2.HOUGH_GRADIENT, 1, 20, param1=th1, param2=th2, minRadius=5, maxRadius=45)
+        if circles is not None:
+            circles=np.int16(np.around(circles))
+            for i in circles[0,:]:
+                if i[2]!=0:
+                    panneau=cv2.resize(image[max(0, i[1]-i[2]):i[1]+i[2], max(0, i[0]-i[2]):i[0]+i[2]], (common.size, common.size))/255
+                    cv2.imshow("panneau", panneau)
+                    prediction=model_panneau(np.array([panneau]), training=False)
+                    print("Prediction:", prediction)
+                    if np.any(np.greater(prediction[0], 0.6)):
+                        id_panneau=np.argmax(prediction[0])
+                        print("   -> C'est un panneau:", tab_panneau[id_panneau], "KM/H")
+                        w, h, c=tab_image_panneau[id_panneau].shape
+                    else:
+                        print("   -> Ce n'est pas un panneau")
+        if id_panneau!=-1:
+            frame[0:h, 0:w, :]=tab_image_panneau[id_panneau]
+        cv2.putText(frame, "fichier:"+video, (30, 30), cv2.FONT_HERSHEY_DUPLEX, 1, (0, 255, 0), 1, cv2.LINE_AA)
+        cv2.imshow("Video", frame)
+        key=cv2.waitKey(1)&0xFF
+        if key==ord('q'):
+            quit()
+        if key==ord('a'):
+            for cpt in range(100):
+                ret, frame=cap.read()
+        if key==ord('f'):
+            break
+
+cv2.destroyAllWindows()

Divers/tutoriel25-3/train_panneau.py

@@ -0,0 +1,140 @@
+import tensorflow as tf
+from tensorflow.keras import layers, models
+import numpy as np
+from sklearn.utils import shuffle
+from sklearn.model_selection import train_test_split
+import cv2
+import os
+import time
+import common
+import dataset
+
+batch_size=128
+nbr_entrainement=20
+
+tab_panneau, tab_image_panneau=common.lire_images_panneaux(common.dir_images_panneaux, common.size)
+
+tab_images=np.array([]).reshape(0, common.size, common.size, 3)
+tab_labels=np.array([]).reshape(0, len(tab_image_panneau))
+
+id=0
+for image in tab_image_panneau:
+    lot=dataset.create_lot_img(image, 12000)
+    tab_images=np.concatenate((tab_images, lot))
+    tab_labels=np.concatenate([tab_labels, np.repeat([np.eye(len(tab_image_panneau))[id]], len(lot), axis=0)])
+    id+=1
+
+
+files=os.listdir(common.dir_images_autres_panneaux)
+if files is None:
+    quit("Le repertoire d'image est vide:".format(common.dir_images_autres_panneaux))
+
+nbr=0
+for file in files:
+    if file.endswith("png"):
+        path=os.path.join(common.dir_images_autres_panneaux, file)
+        image=cv2.resize(cv2.imread(path), (common.size, common.size), cv2.INTER_LANCZOS4)
+        lot=dataset.create_lot_img(image, 700)
+        tab_images=np.concatenate([tab_images, lot])
+        nbr+=len(lot)
+
+tab_labels=np.concatenate([tab_labels, np.repeat([np.full(len(tab_image_panneau), 0)], nbr, axis=0)])
+
+nbr_np=int(len(tab_images)/2)
+
+id=1
+nbr=0
+tab=[]
+for cpt in range(nbr_np):
+    file=common.dir_images_sans_panneaux+"/{:d}.png".format(id)
+    if not os.path.isfile(file):
+        break
+    image=cv2.resize(cv2.imread(file), (common.size, common.size))
+    tab.append(image)
+    id+=1
+    nbr+=1
+
+tab_images=np.concatenate([tab_images, tab])
+tab_labels=np.concatenate([tab_labels, np.repeat([np.full(len(tab_image_panneau), 0)], nbr, axis=0)])
+
+tab_panneau=np.array(tab_panneau)
+tab_images=np.array(tab_images, dtype=np.float32)/255
+tab_labels=np.array(tab_labels, dtype=np.float32) #.reshape([-1, 1])
+
+train_images, test_images, train_labels, test_labels=train_test_split(tab_images, tab_labels, test_size=0.10)
+
+train_ds=tf.data.Dataset.from_tensor_slices((train_images, train_labels)).batch(batch_size)
+test_ds=tf.data.Dataset.from_tensor_slices((test_images, test_labels)).batch(batch_size)
+
+print("train_images", len(train_images))
+print("test_images", len(test_images))
+
+@tf.function
+def train_step(images, labels):
+  with tf.GradientTape() as tape:
+    predictions=model_panneau(images)
+    loss=my_loss(labels, predictions)
+  gradients=tape.gradient(loss, model_panneau.trainable_variables)
+  optimizer.apply_gradients(zip(gradients, model_panneau.trainable_variables))
+  train_loss(loss)
+  train_accuracy(labels, predictions)
+
+def train(train_ds, nbr_entrainement):
+  for entrainement in range(nbr_entrainement):
+    start=time.time()
+    for images, labels in train_ds:
+      train_step(images, labels)
+    message='Entrainement {:04d}: loss: {:6.4f}, accuracy: {:7.4f}%, temps: {:7.4f}'
+    print(message.format(entrainement+1,
+                         train_loss.result(),
+                         train_accuracy.result()*100,
+                         time.time()-start))
+    train_loss.reset_states()
+    train_accuracy.reset_states()
+    test(test_ds)
+
+def my_loss(labels, preds):
+    labels_reshape=tf.reshape(labels, (-1, 1))
+    preds_reshape=tf.reshape(preds, (-1, 1))
+    result=loss_object(labels_reshape, preds_reshape)
+    return result
+    
+def test(test_ds):
+  start=time.time()
+  for test_images, test_labels in test_ds:
+    predictions=model_panneau(test_images)
+    t_loss=my_loss(test_labels, predictions)
+    test_loss(t_loss)
+    test_accuracy(test_labels, predictions)
+  message='   >>> Test: loss: {:6.4f}, accuracy: {:7.4f}%, temps: {:7.4f}'
+  print(message.format(test_loss.result(),
+                       test_accuracy.result()*100,
+                       time.time()-start))
+  test_loss.reset_states()
+  test_accuracy.reset_states()
+
+optimizer=tf.keras.optimizers.Adam()
+loss_object=tf.keras.losses.BinaryCrossentropy()
+train_loss=tf.keras.metrics.Mean()
+train_accuracy=tf.keras.metrics.BinaryAccuracy()
+test_loss=tf.keras.metrics.Mean()
+test_accuracy=tf.keras.metrics.BinaryAccuracy()
+model_panneau=common.panneau_model(len(tab_panneau))
+checkpoint=tf.train.Checkpoint(model_panneau=model_panneau)
+
+print("Entrainement")
+train(train_ds, nbr_entrainement)
+checkpoint.save(file_prefix="./training_panneau/panneau")
+
+quit()
+
+for i in range(len(test_images)):
+    prediction=model_panneau(np.array([test_images[i]]))
+    print("prediction", prediction[0])
+    if np.sum(prediction[0])<0.6:
+        print("Ce n'est pas un panneau")
+    else:
+        print("C'est un panneau:", tab_panneau[np.argmax(prediction[0])])
+    cv2.imshow("image", test_images[i])
+    if cv2.waitKey()&0xFF==ord('q'):
+        break

Divers/tutoriel25/README.md

@@ -0,0 +1,8 @@
+# Tutoriel 25
+## Lecture des panneaux de limitation de vitesse
+
+Les vidéos de ce tutoriel sont disponibles aux adresses suivantes:<br>
+partie 1: https://www.youtube.com/watch?v=PvD5POjXw8Q <br>
+partie 2: https://www.youtube.com/watch?v=TYDi0SNCUr0 <br>
+partie 3: https://www.youtube.com/watch?v=fBysd-Y17Tw
+

Divers/tutoriel25/common.py

@@ -0,0 +1,30 @@
+import tensorflow as tf
+from tensorflow.keras import layers, models
+import os
+import cv2
+
+size=42
+dir_images_panneaux="images_panneaux"
+dir_images_autres_panneaux="images_autres_panneaux"
+dir_images_sans_panneaux="images_sans_panneaux"
+
+def lire_images_panneaux(dir_images_panneaux, size=None):
+    tab_panneau=[]
+    tab_image_panneau=[]
+
+    if not os.path.exists(dir_images_panneaux):
+        quit("Le repertoire d'image n'existe pas: {}".format(dir_images_panneaux))
+
+    files=os.listdir(dir_images_panneaux)
+    if files is None:
+        quit("Le repertoire d'image est vide: {}".format(dir_images_panneaux))
+
+    for file in sorted(files):
+        if file.endswith("png"):
+            tab_panneau.append(file.split(".")[0])
+            image=cv2.imread(dir_images_panneaux+"/"+file)
+            if size is not None:
+                image=cv2.resize(image, (size, size), cv2.INTER_LANCZOS4)
+            tab_image_panneau.append(image)
+            
+    return tab_panneau, tab_image_panneau

Divers/tutoriel25/dataset.py

@@ -0,0 +1,82 @@
+import numpy as np
+import cv2
+from multiprocessing import Pool
+import multiprocessing
+import random
+
+def bruit(image_orig):
+    h, w, c=image_orig.shape
+    n=np.random.randn(h, w, c)*random.randint(5, 30)
+    return np.clip(image_orig+n, 0, 255).astype(np.uint8)
+
+def change_gamma(image, alpha=1.0, beta=0.0):
+    return np.clip(alpha*image+beta, 0, 255).astype(np.uint8)
+    
+def modif_img(img):
+    h, w, c=img.shape
+
+    r_color=[np.random.randint(255), np.random.randint(255), np.random.randint(255)]
+    img=np.where(img==[142, 142, 142], r_color, img).astype(np.uint8)
+
+    if np.random.randint(3):
+        k_max=3
+        kernel_blur=np.random.randint(k_max)*2+1
+        img=cv2.GaussianBlur(img, (kernel_blur, kernel_blur), 0)
+
+    M=cv2.getRotationMatrix2D((int(w/2), int(h/2)), random.randint(-10, 10), 1)
+    img=cv2.warpAffine(img, M, (w, h))
+        
+    if np.random.randint(2):
+        a=int(max(w, h)/5)+1
+        pts1=np.float32([[0, 0], [w, 0], [0, h], [w, h]])
+        pts2=np.float32([[0+random.randint(-a, a), 0+random.randint(-a, a)], [w-random.randint(-a, a), 0+random.randint(-a, a)], [0+random.randint(-a, a), h-random.randint(-a, a)], [w-random.randint(-a, a), h-random.randint(-a, a)]])        
+        M=cv2.getPerspectiveTransform(pts1,pts2)
+        img=cv2.warpPerspective(img, M, (w, h))
+        
+    if np.random.randint(2):
+        r=random.randint(0, 5)
+        h2=int(h*0.9)
+        w2=int(w*0.9)
+        if r==0:
+            img=img[0:w2, 0:h2]
+        elif r==1:
+            img=img[w-w2:w, 0:h2]
+        elif r==2:
+            img=img[0:w2, h-h2:h]
+        elif r==3:
+            img=img[w-w2:w, h-h2:h]
+        img=cv2.resize(img, (h, w))
+
+    if np.random.randint(2):
+        r=random.randint(1, int(max(w, h)*0.15))
+        img=img[r:w-r, r:h-r]
+        img=cv2.resize(img, (h, w))
+
+    if not np.random.randint(4):
+        t=np.empty((h, w, c) , dtype=np.float32)
+        for i in range(h):
+            for j in range(w):
+                for k in range(c):
+                    t[i][j][k]=(i/h)
+        M=cv2.getRotationMatrix2D((int(w/2), int(h/2)), np.random.randint(4)*90, 1)
+        t=cv2.warpAffine(t, M, (w, h))
+        img=(cv2.multiply((img/255).astype(np.float32), t)*255).astype(np.uint8)
+
+    img=change_gamma(img, random.uniform(0.6, 1.0), -np.random.randint(50))
+
+    if not np.random.randint(4):
+        p=(15+np.random.randint(10))/100
+        img=(img*p+50*(1-p)).astype(np.uint8)+np.random.randint(100)
+
+    img=bruit(img)
+        
+    return img
+
+def create_lot_img(image, nbr, nbr_thread=None):
+    if nbr_thread is None:
+        nbr_thread=multiprocessing.cpu_count()
+    lot_original=np.repeat([image], nbr, axis=0)
+    with Pool(nbr_thread) as p:
+        lot_result=p.map(modif_img, lot_original)
+        p.close()
+    return lot_result

Divers/tutoriel25/extract_panneau.py

@@ -0,0 +1,46 @@
+import cv2
+import os
+import numpy as np
+import random
+import common
+
+video_dir="D:\dashcam Cedric"
+
+l=os.listdir(video_dir)
+
+for video in l:
+    if not video.endswith("mp4"):
+        continue
+    cap=cv2.VideoCapture(video_dir+"/"+video)
+
+    print("video:", video)
+    while True:
+        ret, frame=cap.read()
+        if ret is False:
+            break
+        f_w, f_h, f_c=frame.shape
+        frame=cv2.resize(frame, (int(f_h/1.5), int(f_w/1.5)))
+
+        image=frame[200:400, 700:1000]
+        cv2.rectangle(frame, (700, 200), (1000, 400), (255, 255, 255), 1)
+
+        gray=cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
+        circles=cv2.HoughCircles(gray, cv2.HOUGH_GRADIENT, 1, 20, param1=30, param2=60, minRadius=5, maxRadius=45)
+        if circles is not None:
+            circles=np.int16(np.around(circles))
+            for i in circles[0,:]:
+                if i[2]!=0:
+                    panneau=cv2.resize(image[max(0, i[1]-i[2]):i[1]+i[2], max(0, i[0]-i[2]):i[0]+i[2]], (common.size, common.size))/255
+                    cv2.imshow("panneau", panneau)
+        cv2.putText(frame, "fichier:"+video, (30, 30), cv2.FONT_HERSHEY_DUPLEX, 1, (0, 255, 0), 1, cv2.LINE_AA)
+        cv2.imshow("Video", frame)
+        key=cv2.waitKey(1)&0xFF
+        if key==ord('q'):
+            quit()
+        if key==ord('a'):
+            for cpt in range(100):
+                ret, frame=cap.read()
+        if key==ord('f'):
+            break
+
+cv2.destroyAllWindows()

Divers/tutoriel25/genere_panneaux.py

@@ -0,0 +1,29 @@
+import numpy as np
+from sklearn.utils import shuffle
+import cv2
+import common
+import dataset
+
+tab_panneau, tab_image_panneau=common.lire_images_panneaux(common.dir_images_panneaux, common.size)
+
+tab_images=np.array([]).reshape(0, common.size, common.size, 3)
+tab_labels=[]
+
+id=0
+for image in tab_image_panneau:
+    lot=dataset.create_lot_img(image, 1000)
+    tab_images=np.concatenate([tab_images, lot])
+    tab_labels=np.concatenate([tab_labels, np.full(len(lot), id)])
+    id+=1
+
+tab_panneau=np.array(tab_panneau)
+tab_images=np.array(tab_images, dtype=np.float32)/255
+tab_labels=np.array(tab_labels).reshape([-1, 1])
+
+tab_images, tab_labels=shuffle(tab_images, tab_labels)
+
+for i in range(len(tab_images)):
+    cv2.imshow("panneau", tab_images[i])
+    print("label", tab_labels[i], "panneau", tab_panneau[int(tab_labels[i])])
+    if cv2.waitKey()&0xFF==ord('q'):
+        quit()

Divers/tutoriel25/houghcircles.py

@@ -0,0 +1,36 @@
+import cv2
+import numpy as np
+
+param1=30
+param2=55
+dp=1.0
+
+cap=cv2.VideoCapture(0)
+
+while True:
+    ret, frame=cap.read()
+    gray=cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
+    circles=cv2.HoughCircles(gray, cv2.HOUGH_GRADIENT, dp, 20, param1=param1, param2=param2, minRadius=10, maxRadius=50)
+    if circles is not None:
+        circles=np.around(circles).astype(np.int32)
+        for i in circles[0, :]:
+            if i[2]!=0:
+                cv2.circle(frame, (i[0], i[1]), i[2], (0, 255, 0), 4)
+    cv2.putText(frame, "[i|k]dp: {:4.2f}  [o|l]param1: {:d}  [p|m]param2: {:d}".format(dp, param1, param2), (10, 40), cv2.FONT_HERSHEY_COMPLEX_SMALL, 1, (255, 0, 0), 1)
+    cv2.imshow("Video", frame)
+    key=cv2.waitKey(1)&0xFF
+    if key==ord('q'):
+        quit()
+    if key==ord('i'):
+        dp=min(10, dp+0.1)
+    if key==ord('k'):
+        dp=max(0.1, dp-0.1)
+    if key==ord('o'):
+        param1=min(255, param1+1)
+    if key==ord('l'):
+        param1=max(1, param1-1)
+    if key==ord('p'):
+        param2=min(255, param2+1)
+    if key==ord('m'):
+        param2=max(1, param2-1)
+cv2.destroyAllWindows()

Divers/tutoriel29/README.md

@@ -0,0 +1,7 @@
+# Tutoriel 29
+## Segmentation d'image avec l'algorithme des K-moyennes
+
+La vidéo de ce tutoriel est disponible l'adresse suivante:<br>
+https://www.youtube.com/watch?v=ytii3XvapRY
+
+

Divers/tutoriel29/ishihara.py

@@ -0,0 +1,46 @@
+import numpy as np
+from matplotlib import pyplot as plt
+from sklearn.cluster import KMeans
+import cv2
+import glob
+
+k=2
+ESPACE="HSV"
+CH=[0, 2]
+size=200
+
+for image in glob.glob('.\images\*.png'):
+    print("Image:", image)
+
+    # Lecture et affichage de l'image
+    img=cv2.imread(image)
+    img=cv2.resize(img, (size, size))
+    cv2.imshow("image", img)
+
+    # Changement d'espace colorimétrique
+    img=cv2.cvtColor(img, eval("cv2.COLOR_BGR2"+ESPACE))
+    X=img[:, :, CH].reshape(img.shape[0]*img.shape[1], len(CH))
+
+    # Graph 2D des couches A et B
+    if len(CH)==2:
+        plt.scatter(X[:,0], X[:,1], s=3)#, marker='+')
+        plt.show()
+
+    # Algorithme K moyennes
+    kmeans=KMeans(n_clusters=k)
+    #kmeans.fit(X)
+    pred=kmeans.fit_predict(X)
+
+    # Graph 2D des couches A et B après utilisation de l'algorithme K moyennes
+    if len(CH)==2:
+        plt.scatter(X[:,0], X[:,1], c=pred, s=3) #10, marker='+')
+        plt.scatter(kmeans.cluster_centers_[:, 0], kmeans.cluster_centers_[:, 1], s=50, c='red')
+        plt.show()
+
+    # Affichage du résultat
+    pred=pred.reshape(img.shape[0], img.shape[1])
+    pred=pred/(k-1)
+    cv2.imshow("kmeans", pred)
+
+    if cv2.waitKey()&0xFF==ord('q'):
+        break

Divers/tutoriel29/ishihara2.py

@@ -0,0 +1,35 @@
+import numpy as np
+from matplotlib import pyplot as plt
+from sklearn.cluster import KMeans
+import cv2
+import sys
+import glob
+
+min_clusters=2
+max_clusters=4
+ESPACES=["YCrCb", "HSV", "LAB"]
+COUCHES=[[1], [0, 2], [1, 2]]
+size=200
+
+for image in glob.glob('.\images\*.png'):
+    print("Image: {} ".format(image), end='')
+    tab=np.zeros([(len(ESPACES))*size, (max_clusters-min_clusters+1)*size], dtype=np.float32)
+    img=cv2.imread(image)
+    cv2.imshow("image", cv2.resize(img, (2*size, 2*size)))
+    img=cv2.resize(img, (size, size))
+
+    for index in range(len(ESPACES)):
+        img2=cv2.cvtColor(img, eval("cv2.COLOR_BGR2"+ESPACES[index]))
+        X=img2[:, :, COUCHES[index]].reshape(img2.shape[0]*img2.shape[1], len(COUCHES[index]))
+        for k in range(min_clusters, max_clusters+1):
+            sys.stdout.write('.')
+            sys.stdout.flush()
+            kmeans=KMeans(n_clusters=k)
+            pred=kmeans.fit_predict(X)
+            pred=pred.reshape(img2.shape[0], img2.shape[1])
+            pred=pred/(k-1)
+            tab[index*size:(index+1)*size, (k-min_clusters)*size:(k-min_clusters)*size+size]=pred
+    sys.stdout.write('\n')
+    cv2.imshow("kmeans", tab)
+    if cv2.waitKey()&0xFF==ord('q'):
+        break

Divers/tutoriel31/README.md

@@ -0,0 +1,5 @@
+# Tutoriel 31
+## K-moyennes: coefficient silhouette
+
+La vidéo de ce tutoriel est disponible à l'adresse suivante: https://www.youtube.com/watch?v=H_AW_lwvdDk
+

Divers/tutoriel31/silouhette.py

@@ -0,0 +1,42 @@
+import numpy as np
+from matplotlib import pyplot as plt
+from matplotlib.figure import Figure
+from matplotlib.backends.backend_agg import FigureCanvas
+from sklearn.cluster import KMeans
+from sklearn.datasets.samples_generator import make_blobs
+import cv2
+
+cluster_std=1.30
+n_samples=300
+X, y=make_blobs(n_samples=n_samples, centers=5, cluster_std=cluster_std)
+
+fig, (ax1, ax2)=plt.subplots(1, 2)
+canvas=FigureCanvas(fig)
+fig.set_size_inches(11, 6)
+
+k=2
+while 1:
+    ax1.cla()
+    ax1.scatter(X[:,0], X[:,1], marker='+', c="#FF0000")
+
+    kmeans=KMeans(n_clusters=k)
+    pred_y=kmeans.fit_predict(X)
+
+    ax2.cla()
+    ax2.scatter(X[:,0], X[:,1], c=pred_y, marker='+')
+    ax2.scatter(kmeans.cluster_centers_[:, 0], kmeans.cluster_centers_[:, 1], s=50, c='#0000FF')
+    canvas.draw()
+
+    img=np.array(canvas.renderer.buffer_rgba())
+    cv2.putText(img, "Nbr cluster={:02d}    [p|m] nbr clusters   [r] reset  [q] quit".format(k), (250, 50), cv2.FONT_HERSHEY_PLAIN, 1, (0, 0, 255), 2)
+
+    cv2.imshow("plot", img)
+    key=cv2.waitKey()&0xFF
+    if key==ord('p'):
+        k=min(99, k+1)
+    if key==ord('m'):
+        k=max(2, k-1)
+    if key==ord('r'):
+        X, y=make_blobs(n_samples=n_samples, centers=5, cluster_std=cluster_std)
+    if key==ord('q'):
+        quit()

Divers/tutoriel31/silouhette2.py

@@ -0,0 +1,43 @@
+import numpy as np
+from matplotlib import pyplot as plt
+from matplotlib.figure import Figure
+from matplotlib.backends.backend_agg import FigureCanvas
+from sklearn.cluster import KMeans
+from sklearn.datasets.samples_generator import make_blobs
+import cv2
+import glob
+
+k=5
+cluster_std=1.30
+n_samples=300
+X, y=make_blobs(n_samples=n_samples, centers=k, cluster_std=cluster_std)
+
+fig, (ax1, ax2)=plt.subplots(1, 2)
+canvas=FigureCanvas(fig)
+fig.set_size_inches(10, 6)
+
+while 1:
+    ax1.cla()
+    ax1.scatter(X[:,0], X[:,1], marker='+', c="#FF0000")
+    ax1.set_title('Données')
+
+    wcss=[]
+    for i in range(1, 11):
+        kmeans=KMeans(n_clusters=i)
+        kmeans.fit(X)
+        wcss.append(kmeans.inertia_)
+
+    ax2.cla()
+    ax2.plot(range(1, 11), wcss, c="#FF0000")
+    ax2.set_title('WCSS pour "elbow method"')
+
+    canvas.draw()
+    img=np.array(canvas.renderer.buffer_rgba())
+    cv2.putText(img, "[r] reset  [q] quit".format(k), (450, 40), cv2.FONT_HERSHEY_PLAIN, 1, (0, 0, 255), 2)
+
+    cv2.imshow("plot", img)
+    key=cv2.waitKey()&0xFF
+    if key==ord('r'):
+        X, y=make_blobs(n_samples=n_samples, centers=k, cluster_std=cluster_std)
+    if key==ord('q'):
+        quit()

Divers/tutoriel31/silouhette3.py

@@ -0,0 +1,57 @@
+import numpy as np
+from matplotlib import pyplot as plt
+from matplotlib.figure import Figure
+from matplotlib.backends.backend_agg import FigureCanvas
+from sklearn.cluster import KMeans
+from sklearn.datasets.samples_generator import make_blobs
+from sklearn.metrics import silhouette_score
+import cv2
+import glob
+
+k=5
+cluster_std=1.30
+n_samples=300
+fig, ((ax1, ax2), (ax3, ax4))=plt.subplots(2, 2)
+canvas=FigureCanvas(fig)
+fig.set_size_inches(12, 8)
+
+X, y=make_blobs(n_samples=n_samples, centers=k, cluster_std=cluster_std)
+
+while 1:
+    ax1.cla()
+    ax1.plot(X[:,0], X[:,1], "+", c="#FF0000")
+    ax1.set_title('Données')
+
+    wcss=[]
+    tab_silhouette=[]
+    for i in range(2, 11):
+        kmeans=KMeans(n_clusters=i)
+        cluster_labels=kmeans.fit_predict(X)
+        wcss.append(kmeans.inertia_)
+        tab_silhouette.append(silhouette_score(X, cluster_labels))
+
+    ax2.cla()
+    ax2.plot(range(2, 11), wcss, c="#FF0000")
+    ax2.set_title('WCSS pour "elbow method"')
+
+    ax3.cla()
+    ax3.plot(range(2, 11), tab_silhouette, c="#FF0000")
+    ax3.set_title('Coefficient silhouette')
+
+    kmeans=KMeans(n_clusters=np.argmax(tab_silhouette)+2)
+    pred_y=kmeans.fit_predict(X)
+    ax4.cla()
+    ax4.scatter(X[:,0], X[:,1], c=pred_y, marker='+')
+    ax4.set_title('Données + centre clusters')
+    ax4.scatter(kmeans.cluster_centers_[:, 0], kmeans.cluster_centers_[:, 1], s=100, c="#0000FF")
+
+    canvas.draw()
+    img=np.array(canvas.renderer.buffer_rgba())
+    cv2.putText(img, "[r] reset  [q] quit".format(k), (450, 40), cv2.FONT_HERSHEY_PLAIN, 1, (0, 0, 255), 2)
+
+    cv2.imshow("plot", img)
+    key=cv2.waitKey()&0xFF
+    if key==ord('r'):
+        X, y=make_blobs(n_samples=n_samples, centers=k, cluster_std=cluster_std)
+    if key==ord('q'):
+        quit()

Divers/tutoriel36-2/KalmanFilter.py

@@ -0,0 +1,115 @@
+import numpy as np
+
+class KalmanFilter(object):
+    def __init__(self, dt, point, box):
+        self.dt=dt
+
+        # Vecteur d'etat initial
+        self.E=np.matrix([[point[0]], [point[1]], [0], [0], [box[0]], [box[1]]])
+
+        # Matrice de transition
+        self.A=np.matrix([[1, 0, self.dt, 0, 0, 0],
+                          [0, 1, 0, self.dt, 0, 0],
+                          [0, 0, 1, 0, 0, 0],
+                          [0, 0, 0, 1, 0, 0],
+                          [0, 0, 0, 0, 1, 0],
+                          [0, 0, 0, 0, 0, 1]])
+
+        # Matrice d'observation, on observe que x et y
+        self.H=np.matrix([[1, 0, 0, 0, 0, 0],
+                          [0, 1, 0, 0, 0, 0],
+                          [0, 0, 0, 0, 1, 0],
+                          [0, 0, 0, 0, 0, 1]])
+
+        v=1E-5
+        #v=1
+        self.Q=np.matrix([[v, 0, 0, 0, 0, 0],
+                          [0, v, 0, 0, 0, 0],
+                          [0, 0, v, 0, 0, 0],
+                          [0, 0, 0, v, 0, 0],
+                          [0, 0, 0, 0, v, 0],
+                          [0, 0, 0, 0, 0, v]])
+
+        v=1E-5
+        #v=1
+        self.R=np.matrix([[v, 0, 0, 0],
+                          [0, v, 0, 0],
+                          [0, 0, v, 0],
+                          [0, 0, 0, v]])
+
+        self.P=np.eye(self.A.shape[1])
+
+    def predict(self):
+        self.E=np.dot(self.A, self.E)
+        # Calcul de la covariance de l'erreur
+        self.P=np.dot(np.dot(self.A, self.P), self.A.T)+self.Q
+        return self.E
+
+    def update(self, z):
+        # Calcul du gain de Kalman
+        S=np.dot(self.H, np.dot(self.P, self.H.T))+self.R
+        K=np.dot(np.dot(self.P, self.H.T), np.linalg.inv(S))
+
+        # Correction / innovation
+        self.E=np.round(self.E+np.dot(K, (z-np.dot(self.H, self.E))))
+        I=np.eye(self.H.shape[1])
+        self.P=(I-(K*self.H))*self.P
+
+        return self.E
+
+class KalmanFilter_old(object):
+    def __init__(self, dt, point, box):
+        self.dt=dt
+
+        # Vecteur d'etat initial
+        self.E=np.matrix([[point[0]], [point[1]], [0], [0], [box[0]], [box[1]], [0], [0]])
+
+        # Matrice de transition
+        self.A=np.matrix([[1, 0, self.dt, 0, 0, 0, 0, 0],
+                          [0, 1, 0, self.dt, 0, 0, 0, 0],
+                          [0, 0, 1, 0, 0, 0, 0, 0],
+                          [0, 0, 0, 1, 0, 0, 0, 0],
+                          [0, 0, 0, 0, 1, 0, self.dt, 0],
+                          [0, 0, 0, 0, 0, 1, 0, self.dt],
+                          [0, 0, 0, 0, 0, 0, 1, 0],
+                          [0, 0, 0, 0, 0, 0, 0, 1]])
+
+        # Matrice d'observation, on observe que x et y
+        self.H=np.matrix([[1, 0, 0, 0, 0, 0, 0, 0],
+                          [0, 1, 0, 0, 0, 0, 0, 0],
+                          [0, 0, 0, 0, 1, 0, 0, 0],
+                          [0, 0, 0, 0, 0, 1, 0, 0]])
+
+        self.Q=np.matrix([[1, 0, 0, 0, 0, 0, 0, 0],
+                          [0, 1, 0, 0, 0, 0, 0, 0],
+                          [0, 0, 1, 0, 0, 0, 0, 0],
+                          [0, 0, 0, 1, 0, 0, 0, 0],
+                          [0, 0, 0, 0, 1, 0, 0, 0],
+                          [0, 0, 0, 0, 0, 1, 0, 0],
+                          [0, 0, 0, 0, 0, 0, 1, 0],
+                          [0, 0, 0, 0, 0, 0, 0, 1]])
+
+        self.R=np.matrix([[1, 0, 0, 0],
+                          [0, 1, 0, 0],
+                          [0, 0, 1, 0],
+                          [0, 0, 0, 1]])
+
+        self.P=np.eye(self.A.shape[1])
+
+    def predict(self):
+        self.E=np.dot(self.A, self.E)
+        # Calcul de la covariance de l'erreur
+        self.P=np.dot(np.dot(self.A, self.P), self.A.T)+self.Q
+        return self.E
+
+    def update(self, z):
+        # Calcul du gain de Kalman
+        S=np.dot(self.H, np.dot(self.P, self.H.T))+self.R
+        K=np.dot(np.dot(self.P, self.H.T), np.linalg.inv(S))
+
+        # Correction / innovation
+        self.E=np.round(self.E+np.dot(K, (z-np.dot(self.H, self.E))))
+        I=np.eye(self.H.shape[1])
+        self.P=(I-(K*self.H))*self.P
+
+        return self.E

Divers/tutoriel36-2/README.md

@@ -0,0 +1,4 @@
+# Tutoriel 36
+## Filtre de Kalman partie 2
+
+La vidéo de ce tutoriel est disponible à l'adresse suivante: https://www.youtube.com/watch?v=pR0TAFWnDdU

Divers/tutoriel36-2/affiche_video_label.py

@@ -0,0 +1,33 @@
+import cv2
+import numpy as np
+from numpy import genfromtxt
+import os
+import glob
+
+datasets="2DMOT2015Labels/train/"
+dataset="PETS09-S2L1"
+
+dir_images=datasets+"/"+dataset+"/img1/"
+fichier_label=datasets+"/"+dataset+"/gt/gt.txt"
+
+if not os.path.exists(fichier_label):
+    print("Le fichier de label n'existe pas ...", fichier)
+    quit()
+
+data=genfromtxt(fichier_label, delimiter=',')
+id_frame=0
+id_objet=0
+
+for image in glob.glob(dir_images+"*.jpg"):
+    frame=cv2.imread(image)
+
+    mask=data[:, 0]==id_frame
+    for d in data[mask, :]:
+        cv2.rectangle(frame, (int(d[2]), int(d[3])), (int(d[2]+d[4]), int(d[3]+d[5])), (0, 255, 0), 2)
+
+    cv2.imshow("frame", frame)
+
+    key=cv2.waitKey(70)&0xFF
+    if key==ord('q'):
+        quit()
+    id_frame+=1

Divers/tutoriel36-2/suivi_multiple.py

@@ -0,0 +1,180 @@
+import cv2
+import numpy as np
+from numpy import genfromtxt
+from KalmanFilter import KalmanFilter
+import math
+import os
+import glob
+
+datasets="2DMOT2015Labels/train"
+dataset="PETS09-S2L1"
+
+distance_mini=500
+rectangle=0
+trace=0
+
+dir_images=datasets+"/"+dataset+"/img1/"
+fichier_label=datasets+"/"+dataset+"/gt/gt.txt"
+
+if not os.path.exists(dir_images):
+    print("Le repertoire n'existe pas ...", dir_images)
+    quit()
+
+if not os.path.exists(fichier_label):
+    print("Le fichier de label n'existe pas ...", fichier)
+    quit()
+
+objets_points=[]
+objets_id=[]
+objets_KF=[]
+objets_historique=[]
+
+def distance(point, liste_points):
+    distances=[]
+    for p in liste_points:
+        distances.append(np.sum(np.power(p-np.expand_dims(point, axis=-1), 2)))
+    return distances
+
+def trace_historique(tab_points, longueur, couleur=(0, 255, 255)):
+    historique=np.array(tab_points)
+    nbr_point=len(historique)
+    longueur=min(nbr_point, longueur)
+    for i in range(nbr_point-1, nbr_point-longueur, -1):
+        cv2.line(frame,
+                 (historique[i-1, 0], historique[i-1, 1]),
+                 (historique[i, 0], historique[i, 1]),
+                 couleur,
+                 2)
+
+data=genfromtxt(fichier_label, delimiter=',')
+id_frame=0
+id_objet=0
+
+start=0
+for image in glob.glob(dir_images+"*.jpg"):
+    frame=cv2.imread(image)
+
+    # Prediction de l'ensemble des objets + affichage
+    for id_obj in range(len(objets_points)):
+        etat=objets_KF[id_obj].predict()
+        etat=np.array(etat, dtype=np.int32)
+        objets_points[id_obj]=np.array([etat[0], etat[1], etat[4], etat[5]])
+        objets_historique[id_obj].append([etat[0], etat[1]])
+        cv2.circle(frame, (etat[0], etat[1]), 5, (0, 0, 255), 2)
+        if rectangle:
+            cv2.rectangle(frame,
+                          (int(etat[0]-etat[4]/2), int(etat[1]-etat[5]/2)),
+                          (int(etat[0]+etat[4]/2), int(etat[1]+etat[5]/2)),
+                          (0, 0, 255),
+                          2)
+        cv2.arrowedLine(frame,
+                        (etat[0], etat[1]),
+                        (etat[0]+3*etat[2], etat[1]+3*etat[3]),
+                        color=(0, 0, 255),
+                        thickness=2,
+                        tipLength=0.2)
+        cv2.putText(frame,
+                    "ID{:d}".format(objets_id[id_obj]),
+                    (int(etat[0]-etat[4]/2), int(etat[1]-etat[5]/2)),
+                    cv2.FONT_HERSHEY_PLAIN,
+                    1.5,
+                    (255, 0, 0),
+                    2)
+
+        if trace:
+            trace_historique(objets_historique[id_obj], 42)
+
+        # Permet de suivre l'ID 0 avec une flèche
+        if objets_id[id_obj]==0:
+            cv2.arrowedLine(frame,
+                            (etat[0], int(etat[1]-etat[5]/2-80)),
+                            (etat[0], int(etat[1]-etat[5]/2-30)),
+                            color=(0, 0, 255),
+                            thickness=5,
+                            tipLength=0.2)
+
+    # Récupération des objets de la frame concernée
+    mask=data[:, 0]==id_frame
+
+    # Affichage des données (rectangle) du detecteur
+    points=[]
+    for d in data[mask, :]:
+        #if np.random.randint(2):
+        if rectangle:
+            cv2.rectangle(frame, (int(d[2]), int(d[3])), (int(d[2]+d[4]), int(d[3]+d[5])), (0, 255, 0), 2)
+        xm=int(d[2]+d[4]/2)
+        ym=int(d[3]+d[5]/2)
+        cv2.circle(frame, (xm, ym), 2, (0, 255, 0), 2)
+        points.append([xm, ym, int(d[4]), int(d[5])])
+
+    # calcul des distances
+    nouveaux_objets=np.ones((len(points)))
+    tab_distances=[]
+    if len(objets_points):
+        for point_id in range(len(points)):
+            distances=distance(points[point_id], objets_points)
+            tab_distances.append(distances)
+
+        tab_distances=np.array(tab_distances)
+        sorted_distances=np.sort(tab_distances, axis=None)
+
+        for d in sorted_distances:
+            if d>distance_mini:
+                break
+            id1, id2=np.where(tab_distances==d)
+            if not len(id1) or not len(id2):
+                continue
+            tab_distances[id1, :]=distance_mini+1
+            tab_distances[:, id2]=distance_mini+1
+            objets_KF[id2[0]].update(np.expand_dims(points[id1[0]], axis=-1))
+            nouveaux_objets[id1]=0
+
+    # Création du filtre de Kalman pour les nouveaux objets
+    for point_id in range(len(points)):
+        if nouveaux_objets[point_id]:
+            print("NOUVEAU", points[point_id])
+            objets_points.append(points[point_id])
+            objets_KF.append(KalmanFilter(0.5, [points[point_id][0], points[point_id][1]], [points[point_id][2], points[point_id][3]]))
+            objets_id.append(id_objet)
+            objets_historique.append([])
+            id_objet+=1
+
+    # Nettoyage ...
+    tab_id=[]
+    for id_point in range(len(objets_points)):
+        if int(objets_points[id_point][0])<-100 or \
+        int(objets_points[id_point][1])<-100 or \
+            objets_points[id_point][0]>frame.shape[1]+100 or \
+            objets_points[id_point][1]>frame.shape[0]+100:
+            print("SUPPRESSION", objets_points[id_point])
+            tab_id.append(id_point)
+
+    for index in sorted(tab_id, reverse=True):
+        del objets_points[index]
+        del objets_KF[index]
+        del objets_id[index]
+        del objets_historique[index]
+
+    cv2.rectangle(frame, (0, 0), (frame.shape[1], 30), (100, 100, 100), cv2.FILLED)
+    message="Frame: {:03d}  Nbr personne: {:d}  nbr filtre: {:d}   [r]Rectangle: {:3}  [t]Trace: {:3}".format(id_frame,
+                                                                                                            len(points),
+                                                                                                            len(objets_points),
+                                                                                                            "ON" if rectangle else "OFF",
+                                                                                                            "ON" if trace else "OFF")
+    cv2.putText(frame,
+                message,
+                (20, 20),
+                cv2.FONT_HERSHEY_PLAIN,
+                1,
+                (255, 255, 255),
+                1)
+
+    cv2.imshow("frame", frame)
+    key=cv2.waitKey(70)&0xFF
+    if key==ord('r'):
+        rectangle=not rectangle
+    if key==ord('t'):
+        trace=not trace
+    if key==ord('q'):
+        quit()
+    id_frame+=1

Divers/tutoriel36/Detector.py

@@ -0,0 +1,33 @@
+import numpy as np
+import cv2
+
+lo=np.array([80, 50, 50])
+hi=np.array([100, 255, 255])
+
+def detect_inrange(image, surface):
+    points=[]
+    image = cv2.cvtColor(image, cv2.COLOR_BGR2HSV)
+    image=cv2.blur(image, (5, 5))
+    mask=cv2.inRange(image, lo, hi)
+    mask=cv2.erode(mask, None, iterations=2)
+    mask=cv2.dilate(mask, None, iterations=2)
+    elements=cv2.findContours(mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)[-2]
+    elements=sorted(elements, key=lambda x:cv2.contourArea(x), reverse=True)
+    for element in elements:
+        if cv2.contourArea(element)>surface:
+            ((x, y), rayon)=cv2.minEnclosingCircle(element)
+            points.append(np.array([int(x), int(y)]))
+        else:
+            break
+
+    return points, mask
+
+def detect_visage(image):
+    face_cascade=cv2.CascadeClassifier("./haarcascade_frontalface_alt2.xml")
+    points=[]
+    gray=cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
+    face=face_cascade.detectMultiScale(gray, scaleFactor=1.2, minNeighbors=3)
+    for x, y, w, h in face:
+        points.append(np.array([int(x+w/2), int(y+h/2)]))
+
+    return points, None