keras搭建VGG、ResNet、GoogleNet InceptionV3实现图像的多分类任务

前一段时间利用深度学习网络做过一些图像分类的任务，所以先找了一些经典的网络进行学习，如VGG、GoogleNet、ResNet，这些网络都是基于卷积操作搭建的网络，主要的差别有两点：内部卷积核大小和网络深度；正是这两点区别，使得网络的分类精度越来越高。这个里面对网络的解释，个人感觉比较通俗易懂，大家可以参考一下。

基于keras实现分类任务

基于keras利用VGG、ResNet、GoogleNet InceptionV3实现图像的分类任务，下面会给出完整代码，但为了熟悉不同整个网络的特点，建议大家自己搭建一下每个分类网络，毕竟利用keras搭建网络还是比较简单的。

# -*- coding: utf-8 -*-
import os
from keras.utils import plot_model
from keras.applications.resnet50 import ResNet50
from keras.applications.vgg19 import VGG19
from keras.applications.inception_v3 import InceptionV3
from keras.layers import Dense,Flatten,GlobalAveragePooling2D
from keras.models import Model,load_model
from keras.optimizers import SGD
from keras.preprocessing.image import ImageDataGenerator
import matplotlib.pyplot as plt
 
class PowerTransferMode:
    #数据准备
    def DataGen(self, dir_path, img_row, img_col, batch_size, is_train):
        if is_train:
            datagen = ImageDataGenerator(rescale=1./255,
                zoom_range=0.25, rotation_range=15.,
                channel_shift_range=25., width_shift_range=0.02, height_shift_range=0.02,
                horizontal_flip=True, fill_mode='constant')
        else:
            datagen = ImageDataGenerator(rescale=1./255)
 
        generator = datagen.flow_from_directory(
            dir_path, target_size=(img_row, img_col),
            batch_size=batch_size,
            #class_mode='binary',
            class_mode='categorical',
            shuffle=is_train)
 
        return generator
 
    #ResNet模型
    def ResNet50_model(self, lr=0.005, decay=1e-6, momentum=0.9, nb_classes=2, img_rows=197, img_cols=197, RGB=True, is_plot_model=False):
        color = 3 if RGB else 1
        base_model = ResNet50(weights='imagenet', include_top=False, pooling=None, input_shape=(img_rows, img_cols, color),
                              classes=nb_classes)
 
        #冻结base_model所有层，这样就可以正确获得bottleneck特征
        for layer in base_model.layers:
            layer.trainable = False
 
        x = base_model.output
        #添加自己的全链接分类层
        x = Flatten()(x)
        #x = GlobalAveragePooling2D()(x)
        #x = Dense(1024, activation='relu')(x)
        predictions = Dense(nb_classes, activation='softmax')(x)
 
        #训练模型
        model = Model(inputs=base_model.input, outputs=predictions)
        sgd = SGD(lr=lr, decay=decay, momentum=momentum, nesterov=True)
        model.compile(loss='categorical_crossentropy', optimizer=sgd, metrics=['accuracy'])
 
        #绘制模型
        if is_plot_model:
            plot_model(model, to_file='resnet50_model.png',show_shapes=True)
 
        return model
 
 
    #VGG模型
    def VGG19_model(self, lr=0.005, decay=1e-6, momentum=0.9, nb_classes=18, img_rows=197, img_cols=197, RGB=True, is_plot_model=False):
        color = 3 if RGB else 1
        base_model = VGG19(weights='imagenet', include_top=False, pooling=None, input_shape=(img_rows, img_cols, color),
                              classes=nb_classes)
 
        #冻结base_model所有层，这样就可以正确获得bottleneck特征
        for layer in base_model.layers:
            layer.trainable = False
 
        x = base_model.output
        #添加自己的全链接分类层
        x = GlobalAveragePooling2D()(x)
        x = Dense(1024, activation='relu')(x)
        x = Dense(1024, activation='relu')(x)
        predictions = Dense(nb_classes, activation='softmax')(x)
 
        #训练模型
        model = Model(inputs=base_model.input, outputs=predictions)
        sgd = SGD(lr=lr, decay=decay, momentum=momentum, nesterov=True)
        model.compile(loss='categorical_crossentropy', optimizer=sgd, metrics=['accuracy'])
 
        # 绘图
        if is_plot_model:
            plot_model(model, to_file='vgg19_model.png',show_shapes=True)
 
        return model
 
    # InceptionV3模型
    def InceptionV3_model(self, lr=0.005, decay=1e-6, momentum=0.9, nb_classes=18, img_rows=197, img_cols=197, RGB=True,
                    is_plot_model=False):
        color = 3 if RGB else 1
        base_model = InceptionV3(weights='imagenet', include_top=False, pooling=None,
                           input_shape=(img_rows, img_cols, color),
                           classes=nb_classes)
 
        # 冻结base_model所有层，这样就可以正确获得bottleneck特征
        for layer in base_model.layers:
            layer.trainable = False
 
        x = base_model.output
        # 添加自己的全链接分类层
        x = GlobalAveragePooling2D()(x)
        x = Dense(1024, activation='relu')(x)
        
        #x = GlobalAveragePooling2D()(x)
        x = Dense(1024, activation='relu')(x)
        predictions = Dense(nb_classes, activation='softmax')(x)
 
        # 训练模型
        model = Model(inputs=base_model.input, outputs=predictions)
        sgd = SGD(lr=lr, decay=decay, momentum=momentum, nesterov=True)
        model.compile(loss='categorical_crossentropy', optimizer=sgd, metrics=['accuracy'])
 
        # 绘图
        if is_plot_model:
            plot_model(model, to_file='inception_v3_model.png', show_shapes=True)
 
        return model
 
    #训练模型
    def train_model(self, model, epochs, train_generator, steps_per_epoch, validation_generator, validation_steps, model_url, is_load_model=False):
        # 载入模型
        if is_load_model and os.path.exists(model_url):
            model = load_model(model_url)
 
        history_ft = model.fit_generator(
            train_generator,
            steps_per_epoch=steps_per_epoch,
            epochs=epochs,
            validation_data=validation_generator,
            validation_steps=validation_steps)
        # 模型保存
        model.save(model_url,overwrite=True)
        return history_ft
 
    # 画图
    def plot_training(self, history):
      acc = history.history['accuracy']
      val_acc = history.history['val_accuracy']
      loss = history.history['loss']
      val_loss = history.history['val_loss']
      epochs = range(len(acc))
      plt.plot(epochs, acc, 'b-')
      plt.plot(epochs, val_acc, 'r')
      plt.title('Training and validation accuracy')
      plt.figure()
      plt.plot(epochs, loss, 'b-')
      plt.plot(epochs, val_loss, 'r-')
      plt.title('Training and validation loss')
      plt.show()
 
 
if __name__ == '__main__':

    #os.environ['CUDA_VISIBLE_DEVICES'] = '-1' 
    image_size = 256
    batch_size = 32
    epo = 1000
    transfer = PowerTransferMode()
    num_train = 14470
    num_test = 3215
    #得到数据
    train_generator = transfer.DataGen('/home/jjin/skin_diagnosis/class_test_3/train/', image_size, image_size, batch_size, True)
    validation_generator = transfer.DataGen("/home/jjin/skin_diagnosis/class_test_3/test/", image_size, image_size, batch_size, False)
 
    #VGG19
    model = transfer.VGG19_model(nb_classes=18, img_rows=image_size, img_cols=image_size, is_plot_model=False)
    history_ft1 = transfer.train_model(model, epo, train_generator, num_train//batch_size, validation_generator, num_test//batch_size, 'vgg19_model_weights.h5',        is_load_model=False)
 
    #ResNet50
    #model = transfer.ResNet50_model(nb_classes=2, img_rows=image_size, img_cols=image_size, is_plot_model=False)
    #history_ft = transfer.train_model(model, 10, train_generator, 600, validation_generator, 60, 'resnet50_model_weights.h5', is_load_model=False)
 
    #InceptionV3
    model = transfer.InceptionV3_model(nb_classes=18, img_rows=image_size, img_cols=image_size, is_plot_model=True)
    # 分多次进行训练，没训练100次，保存一下模型
    for _ in range(10):   
      history_ft2 = transfer.train_model(model, 100, train_generator, num_train//batch_size, validation_generator, num_test//batch_size,                                 'inception_v3_model_weights      .h5', is_load_model=False)
    # 训练的acc_loss图
    transfer.plot_training(history_ft1)
    transfer.plot_training(history_ft2)

在这里有几点要提醒一下，虽然3个网络都搭建出来了，但我只训练了其中的两个网络，其中，在训练InceptionV3时，我把训练过程分为了10个循环，每个循环的epoch是100，这是为了每一个循环后都能保存一下模型，而不至于因为某些原因，导致训练中断，模型没有保存下来。