Initial commit

Tensorflow/tutoriel32/README.md

@@ -0,0 +1,5 @@
+# Tutoriel Keras
+## Classer de l'information avec keras (exemple avec le cancer de la peau)
+
+La vidéo de ce tutoriel est disponible à l'adresse suivante: https://www.youtube.com/watch?v=_X_yOC7zh4c
+

Tensorflow/tutoriel32/model.py

@@ -0,0 +1,43 @@
+from tensorflow.keras import layers, models
+
+# Fonction d'activation à tester: sigmoid, tanh, relu, 
+
+def model(nbr_sortie, nbr_cc):
+    entree=layers.Input(shape=(75, 100, 3), dtype='float32')
+
+    result=layers.Conv2D(nbr_cc, 5, activation='relu', padding='same')(entree)
+    result=layers.Conv2D(nbr_cc, 5, activation='relu', padding='same')(result)
+    result=layers.BatchNormalization()(result)
+    result=layers.MaxPool2D()(result)
+
+    result=layers.Conv2D(2*nbr_cc, 3, activation='relu', padding='same')(result)
+    result=layers.Conv2D(2*nbr_cc, 3, activation='relu', padding='same')(result)
+    result=layers.BatchNormalization()(result)
+    result=layers.Conv2D(2*nbr_cc, 3, activation='relu', padding='same')(result)
+    result=layers.Conv2D(2*nbr_cc, 3, activation='relu', padding='same')(result)
+    result=layers.BatchNormalization()(result)
+    result=layers.MaxPool2D()(result)
+
+    result=layers.Conv2D(4*nbr_cc, 3, activation='relu', padding='same')(result)
+    result=layers.Conv2D(4*nbr_cc, 3, activation='relu', padding='same')(result)
+    result=layers.BatchNormalization()(result)
+    result=layers.Conv2D(4*nbr_cc, 3, activation='relu', padding='same')(result)
+    result=layers.Conv2D(4*nbr_cc, 3, activation='relu', padding='same')(result)
+    result=layers.BatchNormalization()(result)
+    result=layers.Conv2D(4*nbr_cc, 3, activation='relu', padding='same')(result)
+    result=layers.Conv2D(4*nbr_cc, 3, activation='relu', padding='same')(result)
+    result=layers.BatchNormalization()(result)
+    result=layers.Conv2D(4*nbr_cc, 3, activation='relu', padding='same')(result)
+    result=layers.Conv2D(4*nbr_cc, 3, activation='relu', padding='same')(result)
+    result=layers.BatchNormalization()(result)
+    result=layers.MaxPool2D()(result)
+    
+    result=layers.Flatten()(result)
+    result=layers.Dense(1024, activation='relu')(result)
+    result=layers.Dense(1024, activation='relu')(result)
+    result=layers.BatchNormalization()(result)
+    sortie=layers.Dense(nbr_sortie, activation='softmax')(result)
+
+    model=models.Model(inputs=entree, outputs=sortie)
+    return model
+

Tensorflow/tutoriel32/predict.py

@@ -0,0 +1,95 @@
+import random
+import tensorflow as tf
+import csv
+import numpy as np
+import cv2
+import model
+import os
+os.environ['CUDA_VISIBLE_DEVICES']='-1'
+
+fichier='ISIC2018_Task3_Training_GroundTruth/ISIC2018_Task3_Training_GroundTruth.csv'
+dir_images='ISIC2018_Task3_Training_Input/'
+
+labels=['Melanoma',
+        'Melanocytic nevus',
+        'Basal cell carcinoma',
+        'Actinic keratosis',
+        'Benign keratosis',
+        'Dermatofibroma',
+        'Vascular lesion']
+
+tab_images=[]
+tab_labels=[]
+
+def rotateImage(image, angle):
+    image_center=tuple(np.array(image.shape[1::-1])/2)
+    rot_mat=cv2.getRotationMatrix2D(image_center, angle, 1.0)
+    result=cv2.warpAffine(image, rot_mat, image.shape[1::-1], flags=cv2.INTER_LINEAR)
+    return result
+
+def bruit(image):
+    h, w, c=image.shape
+    n=np.random.randn(h, w, c)*random.randint(5, 30)
+    return np.clip(image+n, 0, 255).astype(np.uint8)
+
+with open(fichier, newline='') as csvfile:
+    lignes=csv.reader(csvfile, delimiter=',')
+    next(lignes, None)
+    for ligne in lignes:
+        label=np.array(ligne[1:], dtype=np.float32)        
+        img=cv2.imread(dir_images+ligne[0]+'.jpg')
+        img=cv2.resize(img, (100, 75))
+        if img is None:
+            print("XXX")
+            quit()
+        tab_labels.append(label)
+        tab_images.append(img)
+
+        if label[1]:
+            continue
+
+        flag=0
+        for angle in range(0, 360, 30):
+            img_r=rotateImage(img, angle)
+            
+            if label[2] or label[3] or label[5] or label[6]:
+                tab_labels.append(label)
+                i=cv2.flip(img_r, 0)
+                tab_images.append(i)
+
+            if not flag%3 and (label[0] or label[4]):
+                tab_labels.append(label)
+                i=cv2.flip(img_r, 0)
+                tab_images.append(i)
+            flag+=1
+                
+            if label[2] or label[3] or label[5] or label[6]:
+                tab_labels.append(label)
+                i=cv2.flip(img_r, 1)
+                tab_images.append(i)
+
+            if label[5] or label[6]:
+                tab_labels.append(label)
+                i=cv2.flip(img_r, -1)
+                tab_images.append(i)
+
+tab_labels=np.array(tab_labels, dtype=np.float32)        
+tab_images=np.array(tab_images, dtype=np.float32)/255
+
+indices=np.random.permutation(len(tab_labels))
+tab_labels=tab_labels[indices]
+tab_images=tab_images[indices]
+
+print("SOMME", np.sum(tab_labels, axis=0))
+
+model=tf.keras.models.load_model('my_model/')
+
+for i in range(len(tab_images)):
+    cv2.imshow("image", tab_images[i])
+    prediction=model.predict(np.array([tab_images[i]], dtype=np.float32))
+    print("Bonne reponse:{}, Reponse du réseau:{}".format(labels[np.argmax(tab_labels[i])], labels[np.argmax(prediction[0])]))
+    key=cv2.waitKey()&0xFF
+    if key==ord('q'):
+        break
+
+cv2.destroyAllWindows()

Tensorflow/tutoriel32/train.py

@@ -0,0 +1,82 @@
+import random
+import tensorflow as tf
+import csv
+import numpy as np
+import cv2
+import model
+
+fichier='ISIC2018_Task3_Training_GroundTruth/ISIC2018_Task3_Training_GroundTruth.csv'
+dir_images='ISIC2018_Task3_Training_Input/'
+
+tab_images=[]
+tab_labels=[]
+
+def rotateImage(image, angle):
+    image_center=tuple(np.array(image.shape[1::-1])/2)
+    rot_mat=cv2.getRotationMatrix2D(image_center, angle, 1.0)
+    result=cv2.warpAffine(image, rot_mat, image.shape[1::-1], flags=cv2.INTER_LINEAR)
+    return result
+
+with open(fichier, newline='') as csvfile:
+    lignes=csv.reader(csvfile, delimiter=',')
+    next(lignes, None)
+    for ligne in lignes:
+        label=np.array(ligne[1:], dtype=np.float32)        
+        img=cv2.imread(dir_images+ligne[0]+'.jpg')
+        if img is None:
+            print("Image absente", dir_images+ligne[0]+'.jpg')
+            quit()
+        img=cv2.resize(img, (100, 75))
+        tab_labels.append(label)
+        tab_images.append(img)
+
+        if label[1]:
+            continue
+
+        flag=0
+        for angle in range(0, 360, 30):
+            img_r=rotateImage(img, angle)
+            
+            if label[2] or label[3] or label[5] or label[6]:
+                tab_labels.append(label)
+                i=cv2.flip(img_r, 0)
+                tab_images.append(i)
+
+            if not flag%3 and (label[0] or label[4]):
+                tab_labels.append(label)
+                i=cv2.flip(img_r, 0)
+                tab_images.append(i)
+            flag+=1
+                
+            if label[2] or label[3] or label[5] or label[6]:
+                tab_labels.append(label)
+                i=cv2.flip(img_r, 1)
+                tab_images.append(i)
+
+            if label[5] or label[6]:
+                tab_labels.append(label)
+                i=cv2.flip(img_r, -1)
+                tab_images.append(i)
+
+tab_labels=np.array(tab_labels, dtype=np.float32)        
+tab_images=np.array(tab_images, dtype=np.float32)/255
+
+indices=np.random.permutation(len(tab_labels))
+tab_labels=tab_labels[indices]
+tab_images=tab_images[indices]
+
+print("SOMME", np.sum(tab_labels, axis=0))
+
+model=model.model(7, 8)
+
+optimizer=tf.keras.optimizers.Adam(learning_rate=1E-4)
+
+model.compile(optimizer=optimizer,
+              loss='categorical_crossentropy',
+              metrics=['accuracy'])
+model.fit(tab_images,
+          tab_labels,
+          validation_split=0.05,
+          batch_size=16,
+          epochs=30)
+model.save('my_model/')