import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import layers
import matplotlib.pyplot as plt
%matplotlib inline
import numpy as np
import glob
import os#（1）创建输入管道
# 导入原始数据
(train_images, train_labels),(_, _) = tf.keras.datasets.mnist.load_data()
# 查看原始数据大小与数据格式
# 60000张图片，每一张图片都是28*28像素
# print(train_images.shape)
# dtype('uint8')，每一位的范围都是0-255的整数，由于图像的一个通道中rgb颜色值就是0-255不等，因此uint8就是图像的标准数字格式
# print(train_images.dtype)#（1.1）数据预处理
# 转换数据类型
train_images = train_images.reshape(train_images.shape[0], 28,28,1)
train_images = train_images.astype('float32')# 归一化0-255>>[-1,1]
train_images = (train_images - 127.5)/127.5#（1.2）确定训练时的BATCH_SIZE与BUFFER_SIZE
BATCH_SIZE = 256 # 每一个batch指一次训练，batch_size表示一次训练所需的数据个数。这里一次训练需要256张图片
BUFFER_SIZE = 60000 # 目前不知道buffer是干什么的#（1.3）将归一化后的图像转化为tf内置的一种数据形式
datasets = tf.data.Dataset.from_tensor_slices(train_images)#（1.4）将训练模型的数据集进行打乱的操作：shuffle
datasets = datasets.shuffle(BUFFER_SIZE).batch(BATCH_SIZE)#（2）生成器模型
def Generator_Model():model = keras.Sequential() # 顺序模型# dense 全连接层# 输入：长度为100的随机数向量（自己定义）# 输出：长度为256的向量model.add(layers.Dense(256, input_shape = (100,), use_bias = False))model.add(layers.BatchNormalization()) # 归一化层model.add(layers.LeakyReLU()) # 激活层model.add(layers.Dense(512, use_bias = False))model.add(layers.BatchNormalization()) # 归一化层model.add(layers.LeakyReLU()) # 激活层model.add(layers.Dense(28*28*1, use_bias = False, activation = 'tanh'))model.add(layers.BatchNormalization()) # 归一化层model.add(layers.Reshape((28,28,1))) # 写为元组的形式return model
#（3）判别器模型
def Discriminator_Model():model = keras.Sequential()model.add(layers.Flatten()) # 将3维图像拉伸为一维图像model.add(layers.Dense(512, use_bias = False))model.add(layers.BatchNormalization()) # 归一化层model.add(layers.LeakyReLU()) # 激活层model.add(layers.Dense(256, use_bias = False))model.add(layers.BatchNormalization()) # 归一化层model.add(layers.LeakyReLU()) # 激活层model.add(layers.Dense(1)) # 输出1或者0return modelcross_entropy = tf.keras.losses.BinaryCrossentropy(from_logits = True)#（4）判别器的损失函数：对于真是图片，判定为1；对于生成图片，判定为0
def discriminator_loss(real_out, fake_out):real_loss = cross_entropy(tf.ones_like(real_out),real_out)fake_loss = cross_entropy(tf.zeros_like(fake_out),fake_out)return real_loss+fake_loss#（5）生成器损失函数：对于生成图片，判定为1
def generator_loss(fake_out):fake_loss = cross_entropy(tf.ones_like(fake_out),fake_out)return fake_loss#（6）创建判别器和生成器的优化器，定义参数的学习速率
generator_opt = tf.keras.optimizers.Adam(1e-4)
discriminator_opt = tf.keras.optimizers.Adam(1e-4)EPOCHS = 100
noise_dim = 100
num_exp_to_generate = 16
seed = tf.random.normal([num_exp_to_generate, noise_dim])# 实例化生成器与判别器
Generator = Generator_Model()
Discriminator = Discriminator_Model()#（7）训练GAN网络
# 每一个batch
def train_step(images):noise = tf.random.normal([BATCH_SIZE, noise_dim])with tf.GradientTape() as gen_tape, tf.GradientTape() as disc_tape:real_output = Discriminator(images, training = True)gen_image = Generator(noise, training = True)fake_output = Discriminator(gen_image, training = True)gen_loss = generator_loss(fake_output)disc_loss = discriminator_loss(real_output, fake_output)#优化gradient_gen = gen_tape.gradient(gen_loss, Generator.trainable_variables)gradient_disc = disc_tape.gradient(disc_loss, Discriminator.trainable_variables)generator_opt.apply_gradients(zip(gradient_gen, Generator.trainable_variables))discriminator_opt.apply_gradients(zip(gradient_disc, Discriminator.trainable_variables))# 可视化函数
def generator_plt_img(gen_model, test_noise):pre_images = gen_model(test_noise, training = False)fig = plt.figure(figsize=(4, 4))for i in range(pre_images.shape[0]):plt.subplot(4,4,i+1)plt.imshow((pre_images[i,:,:,0]+1)/2, cmap = 'gray')plt.axis('off')plt.show()# 完整的训练模型的函数
def train(dataset, epochs):for epoch in range(epochs):for img_batch in dataset:train_step(img_batch)print('.',end='')generator_plt_img(Generator, seed)
# 训练模型
train(datasets, EPOCHS)视频链接：https://www.bilibili.com/video/BV1f7411E7wU/?spm_id_from=333.999.0.0
