resnet changes; option to disable cam training

network/Resnet50.py

@@ -33,9 +33,14 @@ class Resnet50(NetworkBase):
         if shouldSave:
             base_model.save('input/resnet_50_base_model.h5')
 
+        #Set the base_network untrainable, except the last Conv Block
+        for layer in base_model.layers[:-36]:
+            layer.trainable = False
 
         x = GlobalAveragePooling2D()(base_model.output)
-        x = Dense(self.NUM_CLASSES)(x)
+        x = Dropout(0.4)(x)
+        x = Dense(128)(x)
+        x = Dropout(0.2)(x)
         out = Dense(self.NUM_CLASSES, activation='softmax',  name='prediction')(x)
 
         if cam:

train_classifier.py

@@ -33,6 +33,7 @@ parser.add_option("--patience", type="int", dest="early_stopping_patience", help
 parser.add_option("--split_size", type="float", dest="split_size", help="Proportion of validation examples out of all test examples. Set as value between 0 and 1.", default=0.5)
 parser.add_option("--width", type="int", dest="width", default=224)
 parser.add_option("--height", type="int", dest="height", default=224)
+parser.add_option("--no_cam_training", action="store_false", dest="cam_training", default=True)
 # parser.add_option("--output_weight_path", dest="output_weight_path", help="Output path for weights.", default='./model_frcnn.hdf5')
 # parser.add_option("--input_weight_path", dest="input_weight_path", help="Input path for weights. If not specified, will try to load default weights provided by keras.")
 
@@ -111,7 +112,7 @@ test_dataset = test_loader.load_dataset()
 
 # Create Network
 network = Resnet50(n_classes, img_width, img_height, channels, base_model_file)
-model = network.get_model(cam=False)
+model = network.get_model(cam=options.cam_training)
 
 #compile the model
 model.compile(optimizer=RMSprop())