model.py

import helper
import json
import os
from keras.layers import Dense, GlobalAveragePooling2D, Lambda, Flatten, Dropout
from keras.layers.convolutional import Convolution2D
from keras.models import Model, Sequential, model_from_json
from keras.callbacks import EarlyStopping

IMSIZE = 80 # Input image size
SAMPLES_PER_EPOCH = 24576 # Number of samples per epoch
NB_EPOCH = 50 # Number of epoch
NB_VAL_SAMPLES = 2048 # Number of validation samples

try:
    # If model file exists in the folder, load the model and weights from files
    with open('model.json', 'r') as jfile:
        model = model_from_json(json.load(jfile))
    model.summary()
    model.compile("adam", "mse")
    model.load_weights('model.h5', by_name=True)
    print("Model and weights are loaded.")

except:
    # If model file does not exist, construct a Xception model
    print("Construct a new model.")
    model = Sequential()
    model.add(Lambda(lambda x: x / 127.5 - 1.0, input_shape=(80, 80, 3)))
    model.add(Convolution2D(24, 5, 5, activation='relu', border_mode='valid', subsample=(2, 2)))
    model.add(Convolution2D(36, 5, 5, activation='relu', border_mode='valid', subsample=(2, 2)))
    model.add(Convolution2D(48, 5, 5, activation='relu', border_mode='valid', subsample=(2, 2)))
    model.add(Convolution2D(64, 3, 3, activation='relu', border_mode='valid', subsample=(1, 1)))
    model.add(Convolution2D(64, 3, 3, activation='relu', border_mode='valid', subsample=(1, 1)))
    model.add(Flatten())
    model.add(Dropout(0.5))
    model.add(Dense(1164, activation='relu'))
    model.add(Dropout(0.5))
    model.add(Dense(100, activation='relu'))
    model.add(Dense(50, activation='relu'))
    model.add(Dense(10, activation='relu'))
    model.add(Dense(1))
    model.summary()
    model.compile("adam", "mse")

# Use Fit Generator for training and validation data epoches
train_gen = helper.batch_generator()
valid_gen = helper.batch_generator()
early_stopping = EarlyStopping(monitor='val_loss',
                               min_delta=0,
                               patience=3)
model.fit_generator(train_gen,
                    samples_per_epoch=SAMPLES_PER_EPOCH,
                    nb_epoch=NB_EPOCH,
                    validation_data=valid_gen,
                    nb_val_samples=NB_VAL_SAMPLES,
                    callbacks=[early_stopping],
                    verbose=1)

# Ask if overwrite the existing model files
if 'model.json' in os.listdir():
    print("Model exists in the folder. Overwrite (y) or Cancel (n)?")
    user_input = input()
    if user_input == "y":
        helper.save_model(model)
        print("Model and weights are overwritten.")
    else:
        print("Model and weights are cancelled.")
else:
    helper.save_model(model)
    print("Model and weights are saved.")