ResNet

ResNet

背景介绍

  ResNet:由华人学者何凯明大神于2015年提出，其主要体现出了残差相连的优势，故简称ResNet，是2015年ILSVRC竞赛的第一名，是一个很好的图像特征提取模型。

ResNet特点

  使用残差块结构，使得网络能够更多获取之前的信息，并且使学习结果对于权重的变化更加敏感
  使用瓶颈结构，先使用1x1的卷积核进行降维，最后再次使用1x1的卷积核升维，可以降低模型的参数量
  Conv Block：作用是改变图像大小，输入和输出的尺寸不同，因此无法直接残差相连，
  Identity Block：作用是增加网络深度，输入和输出的尺寸相同，可以直接残差相连

不同尺寸ResNet网络结构

ResNet50图像分析

TensorFlow2.0实现

from functools import reduce
import tensorflow.keras as keras


def compose(*funcs):
    if funcs:
        return reduce(lambda f, g: lambda *a, **kw: g(f(*a, **kw)), funcs)
    else:
        raise ValueError('Composition of empty sequence not supported.')


class Conv_Bn_Relu(keras.layers.Layer):
    def __init__(self, filters, kernel_size, strides, padding, name):
        super(Conv_Bn_Relu, self).__init__()
        self._name = name
        self.conv = keras.layers.Conv2D(filters, kernel_size, strides, padding)
        self.bn = keras.layers.BatchNormalization()
        self.relu = keras.layers.ReLU()

    def call(self, inputs, **kwargs):
        conv = self.conv(inputs)
        bn = self.bn(conv)
        output = self.relu(bn)

        return output


class Conv_Bn(keras.layers.Layer):
    def __init__(self, filters, kernel_size, strides, padding, name):
        super(Conv_Bn, self).__init__()
        self._name = name
        self.conv = keras.layers.Conv2D(filters, kernel_size, strides, padding)
        self.bn = keras.layers.BatchNormalization()

    def call(self, inputs, **kwargs):
        conv = self.conv(inputs)
        output = self.bn(conv)

        return output


def res_block(x, filters, strides, type, name):
    shortcut = x
    x = compose(Conv_Bn_Relu(filters // 4, (1, 1), (1, 1), padding='same', name='{}{}_conv_bn_relu1'.format(type, name)),
                Conv_Bn_Relu(filters // 4, (3, 3), strides, padding='same', name='{}{}_conv_bn_relu2'.format(type, name)),
                Conv_Bn(filters, (1, 1), (1, 1), padding='same', name='{}{}_conv_bn3'.format(type, name)))(x)
    if type == 'conv_block':
        shortcut = keras.layers.Conv2D(filters, (1, 1), strides, name='{}{}_shortcut'.format(type, name))(shortcut)
    x = keras.layers.Add(name='{}{}_add'.format(type, name))([x, shortcut])
    x = keras.layers.ReLU(name='{}{}_relu3'.format(type, name))(x)

    return x


def ResNet50(input_shape):
    input_tensor = keras.layers.Input(input_shape, name='input')
    x = input_tensor
    x = compose(keras.layers.ZeroPadding2D((3, 3), name='zeropadding'),
                Conv_Bn_Relu(64, (7, 7), (2, 2), padding='valid', name='conv_bn_relu'),
                keras.layers.MaxPool2D((3, 3), (2, 2), padding='same', name='maxpool'))(x)
    filters = [256, 512, 1024, 2048]
    strides = [(1, 1), (2, 2), (2, 2), (2, 2)]
    times = [3, 4, 6, 3]
    for i in range(len(times)):
        x = res_block(x, filters[i], strides[i], 'conv_block', i + 1)
        for j in range(times[i] - 1):
            x = res_block(x, filters[i], (1, 1), 'identity_block{}_'.format(i + 1), j + 1)
    x = compose(keras.layers.GlobalAveragePooling2D(name='global_averagepool'),
                keras.layers.Dense(1000, activation='softmax', name='dense'))(x)
    model = keras.Model(input_tensor, x, name='ResNet50')

    return model


if __name__ == '__main__':

    model = resnet50(input_shape=(224, 224, 3))
    model.build(input_shape=(None, 224, 224, 3))
    model.summary()

ResNet小结

  ResNet是一种非常有效的特征提取网络，由于减少了Dense层的数量，因此参数量相比于VGG大大减少，参数量只有25M，因此实际任务经常使用。