1.  背景

主要介绍LeNet网络预测在CIFAR-10图像数据集上的训练及预测。

2. CIFAR-10图像数据集简介

        CIFAR-10是一个包含了6W张32*32像素的三通道彩色图像数据集，图像划分为10大类，每个类别包含了6K张图像。其中训练集5W张，测试集1W张。

数据加载及预处理：

def load_and_proc_data():
    (X_train, y_train), (X_test, y_test) = cifar10.load_data()
    print('X_train shape', X_train.shape)
    # X_train shape (50000, 32, 32, 3)
    print(X_train.shape[0], 'train samples')
    print(X_test.shape[0], 'test samples')

    X_train = X_train.astype('float32')
    X_test = X_test.astype('float32')
    X_train /= 255
    X_test /= 255

    # 将类向量转换成二值类别矩阵
    y_train = np_utils.to_categorical(y_train, NB_CLASSES)
    y_test = np_utils.to_categorical(y_test, NB_CLASSES)
    return X_train, X_test, y_train, y_test

3. LeNet网络模型定义

3.1 单层卷积网络

from keras.models import Sequential
from keras.layers.convolutional import Conv2D, MaxPooling2D
from keras.layers.core import Activation, Flatten, Dense, Dropout
from keras.datasets import cifar10
from keras.utils import np_utils
from keras.optimizers import RMSprop

class LeNet:
    @staticmethod
    def build(input_shape, classes):
        model = Sequential()
        model.add(Conv2D(32, kernel_size=3, padding='same', input_shape=input_shape))
        model.add(Activation('relu'))
        model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))
        model.add(Dropout(0.25))

        model.add(Flatten())
        model.add(Dense(512))
        model.add(Activation('relu'))
        model.add(Dropout(0.5))

        model.add(Dense(classes))
        model.add(Activation('softmax'))
        model.summary()  # 概要汇总网络
        return model

3.2 模型结构及相关参数

  3.3 增加模型深度（多层卷积）

class LeNet:
    @staticmethod
    def build(input_shape, classes):
        model = Sequential()
        model.add(Conv2D(32, kernel_size=3, padding='same', input_shape=input_shape))
        model.add(Activation('relu'))
        model.add(Conv2D(32, kernel_size=3, padding='same'))
        model.add(Activation('relu'))
        model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))
        model.add(Dropout(0.25))

        model.add(Conv2D(64, kernel_size=3, padding='same'))
        model.add(Activation('relu'))
        model.add(Conv2D(64, kernel_size=3, padding='same'))
        model.add(Activation('relu'))
        model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))
        model.add(Dropout(0.25))
        
        model.add(Flatten())
        model.add(Dense(512))
        model.add(Activation('relu'))
        model.add(Dropout(0.5))

        model.add(Dense(classes))
        model.add(Activation('softmax'))
        model.summary()  # 概要汇总网络
        return model

4. 模型训练及预测

def model_train(X_train, y_train):
    OPTIMIZER = RMSprop()
    model = LeNet.build(input_shape=INPUT_SHAPE, classes=NB_CLASSES)
    model.compile(loss='categorical_crossentropy', optimizer=OPTIMIZER, metrics=['accuracy'])
    history = model.fit(X_train, y_train, batch_size=BATCH_SIZE, epochs=NB_EPOCH, verbose=1, validation_split=VALIDATION_SPLIT)
    # plot_picture(history)
    return model

def model_evaluate(model, X_test, y_test):
    score = model.evaluate(X_test, y_test, batch_size=BATCH_SIZE, verbose=1)
    print('Test score: ', score[0])
    print('Test acc: ', score[1])

5. 打印准确率和损失函数

import matplotlib.pyplot as plt

def plot_picture(history):
    print(history.history.keys())
    # -----------acc---------------
    plt.plot(history.history['accuracy'])
    plt.plot(history.history['val_accuracy'])
    plt.title('model acc')
    plt.ylabel('acc')
    plt.xlabel('epoch')
    plt.legend(['train', 'test'], loc='upper left')
    plt.show()

    # -----------loss---------------
    plt.plot(history.history['loss'])
    plt.plot(history.history['val_loss'])
    plt.title('model loss')
    plt.ylabel('loss')
    plt.xlabel('epoch')
    plt.legend(['train', 'test'], loc='upper left')
    plt.show()

 6. 模型保存

def model_save(model):
    # 保存网络结构
    model_json = model.to_json()
    with open('cifar10_architecture.json', 'w') as f:
        f.write(model_json)
    # 保存网络权重
    model.save_weights('cifar10_weights.h5', overwrite=True)

7. 主函数

NB_EPOCH = 50
BATCH_SIZE = 128
VALIDATION_SPLIT = 0.2
IMG_ROWS, IMG_COLS = 32, 32
IMG_CHANNELS = 3
INPUT_SHAPE = (IMG_ROWS, IMG_COLS, IMG_CHANNELS)  # 注意顺序
NB_CLASSES = 10

if __name__ == '__main__':
    X_train, X_test, y_train, y_test = load_and_proc_data()
    model = model_train(X_train, y_train)
    # model_save(model)
    model_evaluate(model, X_test, y_test)

模型输出

Test score:  1.3542113304138184
Test acc:  0.6733999848365784

8. 模型加载及在线推理

模型训练好以后，从模型文件加载模型，并进行预测。

import numpy as np
from keras.models import model_from_json
from keras.optimizers import SGD
from skimage.transform import resize
import imageio

def input_data_proc():
    img_names = ['cat.png', 'dog.png']
    img_list = []
    for img_name in img_names:
        img = imageio.imread(img_name)
        img = resize(img, output_shape=(32, 32, 3)).astype('float32')
        print('size: ', img.shape)
        img_list.append(img)
    img_list = np.array(img_list) / 255
    return img_list

def model_predict(model, optim, img_list):
    model.compile(loss='categorical_crossentropy', optimizer=optim, metrics=['accuracy'])
    preds = model.predict(img_list)
    preds = np.argmax(preds, axis=1)
    print(preds)

if __name__ == '__main__':
        model_json = 'cifar10_architecture.json'
        model_weight = 'cifar10_weights.h5'
        model = model_from_json(open(model_json).read())
        model.load_weights(model_weight)

        optim = SGD()
        img_list = input_data_proc()
        model_predict(model, optim, img_list)