fintune inception

binary_image_classifier/simple_keras.py

@@ -3,7 +3,7 @@
 from keras.models import Sequential
 from keras.preprocessing.image import ImageDataGenerator, load_img, img_to_array
 
-TRAIN_PATH = '/Users/sai/dev/catsdogs/data2/train/'
+TRAIN_PATH = '/Users/sai/dev/datasets/catsdogs-kaggle/data2/train/'
 
 # Constants
 NUM_CHANNELS = 3

finetune/inception.py

@@ -1,33 +1,79 @@
 from keras.applications.inception_v3 import InceptionV3, preprocess_input, decode_predictions
 from keras.layers import Dense, GlobalAveragePooling2D
 from keras.models import Model
+from keras.preprocessing.image import ImageDataGenerator
 
-NUM_CLASSES = 2
+TRAIN_PATH = '/Users/sai/dev/datasets/catsdogs-kaggle/data2/train/'
 
+# Constants
+NUM_CHANNELS = 3
+IMG_X = 150
+IMG_Y = 150
 
-def batch_generator():
-    pass
+BATCH_SIZE = 16
+TOTAL_NUM_IMAGES = 25000
 
-# base pre-trained model
-base_model = InceptionV3(include_top=False, weights='imagenet')
 
-# Global
-x = base_model.output
-x = GlobalAveragePooling2D()(x)
+def get_train_data_augmenter():
+    # real time image augmentation
+    augmenter = ImageDataGenerator(
+        # rotation_range=40,
+        # width_shift_range=0.2,
+        # height_shift_range=0.2,
+        shear_range=0.2,
+        zoom_range=0.2,
+        horizontal_flip=True,
+        fill_mode='nearest',
+        # rescale=1./255,
+        preprocessing_function=preprocess_input
+    )
+    return augmenter
 
-# Fully connected layer
-x = Dense(units=1024, activation='relu')(x)
-# Logistic softmax layer
-predictions = Dense(NUM_CLASSES, activation='softmax')(x)
 
-model = Model(inputs=base_model.input, outputs=predictions)
+def get_train_data_generator(augmenter):
+    train_generator = augmenter.flow_from_directory(
+        TRAIN_PATH,
+        target_size=(IMG_X, IMG_Y),
+        batch_size=BATCH_SIZE,
+        class_mode='binary'
+    )
+    return train_generator
 
-# Train only the top layers
-for layer in base_model.layers:
-    layer.trainable = False
 
-# Compile the model
-model.compile(optimizer='rmsprop', loss='categorical_crossentropy')
+def get_model():
+    # base pre-trained model
+    base_model = InceptionV3(include_top=False, weights='imagenet')
 
-# Train the model
-model.fit_generator()
+    # Global
+    x = base_model.output
+    x = GlobalAveragePooling2D()(x)
+
+    # Fully connected layer
+    x = Dense(units=1024, activation='relu')(x)
+    # Logistic softmax layer
+    predictions = Dense(1, activation='sigmoid')(x)
+
+    model = Model(inputs=base_model.input, outputs=predictions)
+
+    # Train only the top layers
+    for layer in base_model.layers:
+        layer.trainable = False
+
+    # Compile the model
+    model.compile(optimizer='rmsprop', loss='binary_crossentropy', metrics=['accuracy'])
+
+    return model
+
+
+def train_model(model, data_gen):
+    model.fit_generator(
+        data_gen,
+        steps_per_epoch=TOTAL_NUM_IMAGES // BATCH_SIZE,
+        epochs=50
+    )
+
+if __name__ == "__main__":
+    augmenter = get_train_data_augmenter()
+    model = get_model()
+    train_data_gen = get_train_data_generator(augmenter)
+    train_model(model, train_data_gen)