vgg12_bn_gap.py

from keras.models import Sequential
from keras.layers.core import Flatten, Dense, Dropout
from keras.layers.convolutional import Convolution2D, MaxPooling2D, ZeroPadding2D
from keras.layers import BatchNormalization, GlobalAveragePooling2D
from keras.optimizers import SGD
import numpy as np
import time

img_size = 128

def VGG_12(weights_path=None):
    model = Sequential()
    model.add(Convolution2D(64, kernel_size=(3, 3), activation='relu', padding='same',
                            input_shape=(img_size,int(img_size/2),3)))
    model.add(BatchNormalization())
    model.add(Convolution2D(64, kernel_size=(3, 3), activation='relu', padding='same'))
    model.add(BatchNormalization())
    
    model.add(Convolution2D(128, kernel_size=(3, 3), activation='relu', strides=(2,2), padding='same'))
    model.add(BatchNormalization())
    model.add(Convolution2D(128, kernel_size=(3, 3), activation='relu', padding='same'))
    model.add(BatchNormalization())
    
    model.add(Convolution2D(256, kernel_size=(3, 3), activation='relu', strides=(2,2), padding='same'))
    model.add(BatchNormalization())
    model.add(Convolution2D(256, kernel_size=(3, 3), activation='relu', padding='same'))
    model.add(BatchNormalization())
    model.add(Convolution2D(256, kernel_size=(3, 3), activation='relu', padding='same'))
    model.add(BatchNormalization())

    model.add(Convolution2D(512, kernel_size=(3, 3), activation='relu', strides=(2,2), padding='same'))
    model.add(BatchNormalization())
    model.add(Convolution2D(512, kernel_size=(3, 3), activation='relu', padding='same'))
    model.add(BatchNormalization())
    model.add(Convolution2D(512, kernel_size=(3, 3), activation='relu', padding='same'))
    model.add(BatchNormalization())

    model.add(GlobalAveragePooling2D())
    model.add(BatchNormalization())
    model.add(Dense(1000, activation='relu'))
    model.add(Dense(1000, activation='softmax'))

    if weights_path:
        model.load_weights(weights_path)

    return model

if __name__ == "__main__":
    im = np.ones((1, img_size, int(img_size/2), 3))

    model = VGG_12()
    sgd = SGD(lr=0.1, decay=1e-6, momentum=0.9, nesterov=True)
    model.compile(optimizer=sgd, loss='categorical_crossentropy')
    for i in range(10):
        start_time = time.time()
        out = model.predict(im)
        print("--- %s seconds ---" % (time.time() - start_time))