def save_bottleneck_features(self, train_data_dir, val_data_dir,
batch_size):
train_generator = util.get_generator(train_data_dir, self.img_width,
self.img_height, batch_size)
val_generator = util.get_generator(val_data_dir, self.img_width,
self.img_height, batch_size)
model = self.get_base_model()
util.save_model_plot(self.base_model_plot_path, model)
train_bottleneck_features = model.predict_generator(
train_generator,
len(train_generator.filenames) // batch_size)
util.save_bottleneck_features(self.train_bottleneck_features_path,
train_bottleneck_features)
val_bottleneck_features = model.predict_generator(
val_generator,
len(val_generator.filenames) // batch_size)
util.save_bottleneck_features(self.val_bottleneck_features_path,
val_bottleneck_features)
def get_wrong_predictions(self,
directory,
batch_size=defaults['batch_size']):
validation_data = util.load_bottleneck_features(
self.val_bottleneck_features_path)
val_generator = util.get_generator(directory, self.img_width,
self.img_height, batch_size)
train_labels = val_generator.classes
top_model = load_model(self.model_path)
return np.nonzero(
top_model.predict_classes(validation_data).reshape((
-1, )) != train_labels)
def train_top_model(self,
train_data_dir,
val_data_dir,
epochs=defaults['epochs'],
batch_size=defaults['batch_size']):
train_generator = util.get_generator(train_data_dir, self.img_width,
self.img_height, batch_size)
val_generator = util.get_generator(val_data_dir, self.img_width,
self.img_height, batch_size)
train_data = util.load_bottleneck_features(
self.train_bottleneck_features_path)
val_data = util.load_bottleneck_features(
self.val_bottleneck_features_path)
num_classes = train_generator.num_classes
train_labels = train_generator.classes
train_labels = to_categorical(train_labels, num_classes=num_classes)
val_labels = val_generator.classes
val_labels = to_categorical(val_labels, num_classes=num_classes)
model = self.get_top_model(train_data.shape[1:], num_classes)
model.compile(optimizer='rmsprop',
loss='categorical_crossentropy',
metrics=['accuracy'])
util.save_model_plot(self.top_model_plot_path, model)
history = model.fit(train_data,
train_labels,
epochs=epochs,
batch_size=batch_size,
validation_data=(val_data, val_labels))
model.save(self.model_path)
util.save_history(self.history_path, history)
util.eval_model_loss_acc(model, val_data, val_labels, batch_size)
def get_wrong_predictions(self,
directory,
batch_size=defaults['batch_size']):
generator = util.get_generator(directory, self.img_width,
self.img_height, batch_size)
true_labels = generator.classes
model = self.get_trained_model()
predictions = model.predict_generator(generator)
pred_labels = np.argmax(predictions, axis=-1)
wrong_pred = []
for true_label, pred_label, index in zip(true_labels, pred_labels,
range(0, len(true_labels))):
if true_label != pred_label:
wrong_pred.append((index, true_label, pred_label))
return wrong_pred
def evaluate(self, data_dir, batch_size=defaults['batch_size']):
test_generator = util.get_generator(data_dir, self.img_width, self.img_height, batch_size)
base_model = self.get_base_model()
test_bottleneck_features = base_model.predict_generator(
test_generator, len(test_generator.filenames) // batch_size)
num_classes = test_generator.num_classes
test_labels = test_generator.classes
test_labels = to_categorical(test_labels, num_classes=num_classes)
top_model = load_model(self.model_path)
test_loss, test_acc = top_model.evaluate(test_bottleneck_features, test_labels, batch_size=batch_size)
print('Test accuracy: ', test_acc)
print('Test loss: ', test_loss)
