Refactoring

cnn.py

@@ -27,16 +27,18 @@
 classifier = Sequential()
 
 classifier.add(Conv2D(64, (5, 5), input_shape = (256, 256, 1), activation = 'relu'))
-classifier.add(Dropout(0.25))
+classifier.add(Dropout(0.2))
 classifier.add(MaxPooling2D(pool_size = (4, 4)))
 
 classifier.add(Conv2D(32, (3, 3), activation = 'relu'))
-classifier.add(Dropout(0.25))
+classifier.add(Dropout(0.2))
 classifier.add(MaxPooling2D(pool_size = (4, 4)))
 
+
+
 classifier.add(Flatten())
 classifier.add(Dense(units = 128, activation = 'relu'))
-classifier.add(Dropout(0.5))
+classifier.add(Dropout(0.3))
 classifier.add(Dense(units = 32, activation = 'relu'))
 classifier.add(Dense(units = 1, activation = 'sigmoid'))
 
@@ -47,17 +49,21 @@
 
 dir = "Data/Processed0"
 for filename in os.listdir(dir):
-    img = np.load(os.path.join(dir, filename))
-    im_rez = skimage.transform.resize(img, (256, 256, 1))
-    X.append(im_rez)
-    y.append(0)
+	img = np.load(os.path.join(dir, filename))[:,:,0]
+	smooth_img = skimage.filters.gaussian(img, 8)
+	sobel_img = skimage.filters.sobel(img)
+	im_rez = skimage.transform.resize(sobel_img, (256, 256, 1))
+	X.append(im_rez)
+	y.append(0)
 
 dir = "Data/Processed1"
 for filename in os.listdir(dir):
-    img = np.load(os.path.join(dir, filename))
-    im_rez = skimage.transform.resize(img, (256, 256, 1))
-    X.append(im_rez)
-    y.append(1)
+	img = np.load(os.path.join(dir, filename))[:,:,0]
+	smooth_img = skimage.filters.gaussian(img, 8)
+	sobel_img = skimage.filters.sobel(img)
+	im_rez = skimage.transform.resize(sobel_img, (256, 256, 1))
+	X.append(im_rez)
+	y.append(1)
 
 X = np.array(X)
 y = np.array(y)
@@ -66,14 +72,14 @@
 
 print(type(Xtr))
 classifier.summary()
-es = EarlyStopping(patience=10, restore_best_weights = True) 
+es = EarlyStopping(patience=20, restore_best_weights = True) 
 
 history = classifier.fit(Xtr, ytr,
                callbacks = [es],
-               epochs = 100,
+               epochs = 40,
                batch_size = 8,
                validation_data = (Xts, yts))
 classifier.evaluate(Xts, yts)
-classifier.save("model_even_better.h5")
+classifier.save("Models/Keras/Test2/model2.h5")
 import pickle as pickle
-pickle.dump(history, open("history_even_better.pkl", "rb"))
+pickle.dump(history, open("Models/Keras/Test2/history2.pkl", "wb"))

finish_CNN.py

@@ -8,7 +8,7 @@
 import pydicom as pd
 import skimage 
 os.environ["CUDA_VISIBLE_DEVICES"] = "4"
-model = load_model("Models/Keras/model_even_better.h5")
+model = load_model("Models/Keras/model_better.h5")
 
 for fn in glob.iglob(os.path.join("/data3/wv2019/data/processed/", "*", "*", "*"), recursive=True):
     if(os.path.isdir(fn)):