resnet101.py

# -*- coding: utf-8 -*-

from keras.models import Sequential
from keras.optimizers import SGD
from keras.layers import Input, Dense, Convolution2D, MaxPooling2D, AveragePooling2D, ZeroPadding2D, Dropout, Flatten, merge, Reshape, Activation
from keras.layers.normalization import BatchNormalization
from keras.models import Model
from keras import backend as K

from sklearn.metrics import log_loss

#from custom_layers.scale_layer import Scale
from scale_layer import Scale
import keras.regularizers

#from load_cifar10 import load_cifar10_data

import sys
sys.setrecursionlimit(3000)

def identity_block(input_tensor, kernel_size, filters, stage, block):
    '''The identity_block is the block that has no conv layer at shortcut
    # Arguments
        input_tensor: input tensor
        kernel_size: defualt 3, the kernel size of middle conv layer at main path
        filters: list of integers, the nb_filters of 3 conv layer at main path
        stage: integer, current stage label, used for generating layer names
        block: 'a','b'..., current block label, used for generating layer names
    '''
    eps = 1.1e-5
    nb_filter1, nb_filter2, nb_filter3 = filters
    conv_name_base = 'res' + str(stage) + block + '_branch'
    bn_name_base = 'bn' + str(stage) + block + '_branch'
    scale_name_base = 'scale' + str(stage) + block + '_branch'

    x = Convolution2D(nb_filter1, 1, 1, name=conv_name_base + '2a', bias=False)(input_tensor)
    x = BatchNormalization(epsilon=eps, axis=bn_axis, name=bn_name_base + '2a')(x)
    x = Scale(axis=bn_axis, name=scale_name_base + '2a')(x)
    x = Activation('relu', name=conv_name_base + '2a_relu')(x)

    x = ZeroPadding2D((1, 1), name=conv_name_base + '2b_zeropadding')(x)
    x = Convolution2D(nb_filter2, kernel_size, kernel_size,
                      name=conv_name_base + '2b', bias=False)(x)
    x = BatchNormalization(epsilon=eps, axis=bn_axis, name=bn_name_base + '2b')(x)
    x = Scale(axis=bn_axis, name=scale_name_base + '2b')(x)
    x = Activation('relu', name=conv_name_base + '2b_relu')(x)

    x = Convolution2D(nb_filter3, 1, 1, name=conv_name_base + '2c', bias=False)(x)
    x = BatchNormalization(epsilon=eps, axis=bn_axis, name=bn_name_base + '2c')(x)
    x = Scale(axis=bn_axis, name=scale_name_base + '2c')(x)

    x = merge([x, input_tensor], mode='sum', name='res' + str(stage) + block)
    x = Activation('relu', name='res' + str(stage) + block + '_relu')(x)
    return x

def conv_block(input_tensor, kernel_size, filters, stage, block, strides=(2, 2)):
    '''conv_block is the block that has a conv layer at shortcut
    # Arguments
        input_tensor: input tensor
        kernel_size: defualt 3, the kernel size of middle conv layer at main path
        filters: list of integers, the nb_filters of 3 conv layer at main path
        stage: integer, current stage label, used for generating layer names
        block: 'a','b'..., current block label, used for generating layer names
    Note that from stage 3, the first conv layer at main path is with subsample=(2,2)
    And the shortcut should have subsample=(2,2) as well
    '''
    eps = 1.1e-5
    nb_filter1, nb_filter2, nb_filter3 = filters
    conv_name_base = 'res' + str(stage) + block + '_branch'
    bn_name_base = 'bn' + str(stage) + block + '_branch'
    scale_name_base = 'scale' + str(stage) + block + '_branch'

    x = Convolution2D(nb_filter1, 1, 1, subsample=strides,
                      name=conv_name_base + '2a', bias=False)(input_tensor)
    x = BatchNormalization(epsilon=eps, axis=bn_axis, name=bn_name_base + '2a')(x)
    x = Scale(axis=bn_axis, name=scale_name_base + '2a')(x)
    x = Activation('relu', name=conv_name_base + '2a_relu')(x)

    x = ZeroPadding2D((1, 1), name=conv_name_base + '2b_zeropadding')(x)
    x = Convolution2D(nb_filter2, kernel_size, kernel_size,
                      name=conv_name_base + '2b', bias=False)(x)
    x = BatchNormalization(epsilon=eps, axis=bn_axis, name=bn_name_base + '2b')(x)
    x = Scale(axis=bn_axis, name=scale_name_base + '2b')(x)
    x = Activation('relu', name=conv_name_base + '2b_relu')(x)

    x = Convolution2D(nb_filter3, 1, 1, name=conv_name_base + '2c', bias=False)(x)
    x = BatchNormalization(epsilon=eps, axis=bn_axis, name=bn_name_base + '2c')(x)
    x = Scale(axis=bn_axis, name=scale_name_base + '2c')(x)

    shortcut = Convolution2D(nb_filter3, 1, 1, subsample=strides,
                             name=conv_name_base + '1', bias=False)(input_tensor)
    shortcut = BatchNormalization(epsilon=eps, axis=bn_axis, name=bn_name_base + '1')(shortcut)
    shortcut = Scale(axis=bn_axis, name=scale_name_base + '1')(shortcut)

    x = merge([x, shortcut], mode='sum', name='res' + str(stage) + block)
    x = Activation('relu', name='res' + str(stage) + block + '_relu')(x)
    return x

def resnet101_model(img_rows, img_cols, color_type=1, num_classes=None):
    """
    Resnet 101 Model for Keras

    Model Schema and layer naming follow that of the original Caffe implementation
    https://github.com/KaimingHe/deep-residual-networks

    ImageNet Pretrained Weights
    Theano: https://drive.google.com/file/d/0Byy2AcGyEVxfdUV1MHJhelpnSG8/view?usp=sharing
    TensorFlow: https://drive.google.com/file/d/0Byy2AcGyEVxfTmRRVmpGWDczaXM/view?usp=sharing

    Parameters:
      img_rows, img_cols - resolution of inputs
      channel - 1 for grayscale, 3 for color
      num_classes - number of class labels for our classification task
    """
    eps = 1.1e-5

    # Handle Dimension Ordering for different backends
    global bn_axis
    if K.image_dim_ordering() == 'tf':
      bn_axis = 3
      img_input = Input(shape=(img_rows, img_cols, color_type), name='data')
    else:
      bn_axis = 1
      img_input = Input(shape=(color_type, img_rows, img_cols), name='data')

    x = ZeroPadding2D((3, 3), name='conv1_zeropadding')(img_input)
    x = Convolution2D(64, 7, 7, subsample=(2, 2), name='conv1', bias=False)(x)
    x = BatchNormalization(epsilon=eps, axis=bn_axis, name='bn_conv1')(x)
    x = Scale(axis=bn_axis, name='scale_conv1')(x)
    x = Activation('relu', name='conv1_relu')(x)
    x = MaxPooling2D((3, 3), strides=(2, 2), name='pool1')(x)

    x = conv_block(x, 3, [64, 64, 256], stage=2, block='a', strides=(1, 1))
    x = identity_block(x, 3, [64, 64, 256], stage=2, block='b')
    x = identity_block(x, 3, [64, 64, 256], stage=2, block='c')

    x = conv_block(x, 3, [128, 128, 512], stage=3, block='a')
    for i in range(1,4):
      x = identity_block(x, 3, [128, 128, 512], stage=3, block='b'+str(i))

    x = conv_block(x, 3, [256, 256, 1024], stage=4, block='a')
    for i in range(1,23):
      x = identity_block(x, 3, [256, 256, 1024], stage=4, block='b'+str(i))

    x = conv_block(x, 3, [512, 512, 2048], stage=5, block='a')
    x = identity_block(x, 3, [512, 512, 2048], stage=5, block='b')
    x = identity_block(x, 3, [512, 512, 2048], stage=5, block='c')

    x_fc = AveragePooling2D((7, 7), name='avg_pool')(x)
    x_fc = Flatten()(x_fc)
    x_fc = Dense(1000, activation='softmax', name='fc1000')(x_fc)

    model = Model(img_input, x_fc)

    if K.image_dim_ordering() == 'th':
      # Use pre-trained weights for Theano backend
      weights_path = 'imagenet_models/resnet101_weights_th.h5'
    else:
      # Use pre-trained weights for Tensorflow backend
      weights_path = '/home/mara/camnet/imagenet_models/resnet101_weights_tf.h5'

    model.load_weights(weights_path, by_name=True)

    # Truncate and replace softmax layer for transfer learning
    # Cannot use model.layers.pop() since model is not of Sequential() type
    # The method below works since pre-trained weights are stored in layers but not in the model
    x_newfc = AveragePooling2D((7, 7), name='avg_pool')(x)
    x_newfc = Flatten()(x_newfc)
    x_newfc = Dense(num_classes, activation='sigmoid', name='fc8', kernel_regularizer=keras.regularizers.l2(0.01))(x_newfc)

    model = Model(img_input, x_newfc)

    # Learning rate is changed to 0.001
    #sgd = SGD(lr=1e-4, decay=1e-3, momentum=0.9, nesterov=True)
    #model.compile(optimizer=sgd, loss='binary_crossentropy', metrics=['accuracy'])

    return model

if __name__ == '__main__':

    # Example to fine-tune on 3000 samples from Cifar10
    '''
    img_rows, img_cols = 224, 224 # Resolution of inputs
    channel = 3
    num_classes = 10
    batch_size = 16
    nb_epoch = 10

    # Load Cifar10 data. Please implement your own load_data() module for your own dataset
    X_train, Y_train, X_valid, Y_valid = load_cifar10_data(img_rows, img_cols)

    # Load our model
    model = resnet101_model(img_rows, img_cols, channel, num_classes)

    # Start Fine-tuning
    model.fit(X_train, Y_train,
              batch_size=batch_size,
              nb_epoch=nb_epoch,
              shuffle=True,
              verbose=1,
              validation_data=(X_valid, Y_valid),
              )

    # Make predictions
    predictions_valid = model.predict(X_valid, batch_size=batch_size, verbose=1)

    # Cross-entropy loss score
    score = log_loss(Y_valid, predictions_valid)'''