remove pre-train

train_classifier.py

@@ -101,49 +101,15 @@ test_dataset = test_loader.load_dataset()
 network = Resnet50(n_classes, img_width, img_height, channels, resnet_file)
 model = network.get_model()
 
-#compile the model
+for layer in model.layers[154:]:
+    layer.trainable = True
 
+#compile the model
 model.compile(optimizer=RMSprop(), loss='categorical_crossentropy', metrics=['accuracy'])
 
 # Print Network summary
 # model.summary()
 
-'''
-    Pre-Train FC Layers 
-'''
-callbacks_prefit = [
-    ModelCheckpoint(
-        filepath= str(checkpoint_path) + '/best_pre_train.hdf5',
-        save_best_only=True,
-        monitor='val_loss',
-        verbose=1),
-    EarlyStopping(monitor='val_loss', patience=10),
-    TensorBoard(options.output_path +'/logs/{}_prefit'.format(time)),
-]
-
-history = model.fit(train_dataset,
-                    epochs=options.num_epochs_pre_train,
-                    validation_data=validation_dataset,
-                    callbacks=callbacks_prefit,
-                    verbose=2
-                    )
-#model.save_weights('last_pre_train_model.h5')
-print('\nHistory dict:', history.history)
-
-'''
-    Run refinement training on best model of pre-train 
-'''
-model = load_model(str(checkpoint_path) + '/best_pre_train.hdf5')
-
-for layer in model.layers[154:]:
-    layer.trainable = True
-
-model.compile(optimizer=SGD(lr=1e-4, momentum=0.9), loss='categorical_crossentropy', metrics=['accuracy'])
-
-# Reload Training Data to shuffle and augment in a different way then before
-train_dataset = train_loader.load_dataset()
-validation_dataset = validation_loader.load_dataset()
-
 callbacks = [
     ModelCheckpoint(
         filepath= str(checkpoint_path) + '/model.{epoch:03d}-{val_loss:.3f}.hdf5',
@@ -166,7 +132,6 @@ history = model.fit(train_dataset,
                     callbacks=callbacks,
                     verbose=2
                     )
-
 print('\nHistory dict:', history.history)