keras 绘制历史训练图像

history_fer2013 = model.fit_generator(train_generator, 
                              steps_per_epoch=len(y_train)//batch_size, 
                              epochs=epochs, 
                              validation_data=valid_generator, 
                              validation_steps=len(y_valid)//batch_size, 
                              callbacks=callback)

保存训练记录

import pickle
 
with open('../train_results/his_cnn2.pkl', 'wb') as file_pi:
    pickle.dump(history.history, file_pi)

读取训练记录

with open('../train_results/his_cnn2.pkl','rb') as file_pi:
    history=pickle.load(file_pi)

可视化训练过程的完整代码

def load_file(filename):
    """

    :param filename:
    :return:
    """
    import pickle
    data_file = open(filename, 'rb')
    data = pickle.load(data_file)
    data_file.close()
    return data.history


def plot_loss(his):
    """

    :param his:
    :return:
    """
    import matplotlib.pyplot as plt
    plt.figure(figsize=(12, 8))
    plt.plot(np.arange(len(his['loss'])), his['loss'], label='train loss')
    plt.plot(np.arange(len(his['val_loss'])), his['val_loss'], label='valid loss')
    plt.title('loss')
    plt.legend(loc='best')
    plt.savefig('../train_results/loss_cnn2.png')
    plt.show()


def plot_acc(his):
    """

    :param his:
    :return:
    """
    import matplotlib.pyplot as plt
    plt.figure(figsize=(12, 8))
    plt.plot(np.arange(len(his['acc'])), his['acc'], label='train accuracy')
    plt.plot(np.arange(len(his['val_acc'])), his['val_acc'], label='valid accuracy')
    plt.title('accuracy')
    plt.legend(loc='best')
    plt.savefig('../train_results/accuracy_cnn2.png')
    plt.show()


if __name__ == '__main__':
    import numpy as np
    history = load_file('../train_results/his_cnn2.pkl')
    print(np.max(history['val_acc']))
    plot_loss(history)
    plot_acc(history)

参考：
keras训练历史记录的保存与读取
深度学习项目-人脸表情识别