对抗神经网络（DCGAN ）tensorflow2.0实现

对抗神经网络python实现

标题 GAN生成对抗式神经网络数学推导

	 可以参考这一篇博客，我觉得写的很不错：
	 https://blog.csdn.net/weixinhum/article/details/85236329

环境：
python3.x
tensorflow 2.3

代码在上面所属环境中都是通过的

#!/usr/bin/env python
#-*- coding:utf-8 -*-
# author:一条有梦想的咸鱼
# datetime:2020/8/24 15:47
# software: PyCharm

import tensorflow as tf
from tensorflow.keras import layers
import time
import matplotlib.pyplot as plt
import os

#创建模型
'''
生成模型：
全连接层：7*7*256；规范化；激活函数LeakyReLU()；layers.Reshape((7, 7, 256))
上采样：layers.Conv2DTranspose(128, (5, 5), strides=(1, 1), padding='same', use_bias=False)；规范化，激活层
上采样：layers.Conv2DTranspose(64, (5, 5), strides=(2, 2), padding='same', use_bias=False)；规范化，激活层
上采样：layers.Conv2DTranspose(1, (5, 5), strides=(2, 2), padding='same', use_bias=False, activation='tanh')
'''

def make_generator_model():
    model=tf.keras.Sequential()
    model.add(layers.Dense((7*7*256),use_bias=False,input_shape=(100,)))
    model.add(layers.BatchNormalization())
    model.add(layers.LeakyReLU())
    model.add(layers.Reshape((7,7,256)))

    model.add(layers.Conv2DTranspose(128,(5,5),strides=(1,1),padding='same',use_bias=False))
    model.add(layers.BatchNormalization())
    model.add(layers.LeakyReLU())

    model.add(layers.Conv2DTranspose(64,(5,5),strides=(2,2),padding='same',use_bias=False))
    model.add(layers.BatchNormalization())
    model.add(layers.LeakyReLU())

    model.add(layers.Conv2DTranspose(1,(5,5),(2,2),padding='same',use_bias=False,activation='tanh'))

    return model



'''
判别模型：
两层卷积，一层平铺，一层全连接
卷积层：64，卷积核（5，5），步长（2，2），激活函数LeakyReLU()，丢失比例0.3
卷积层：128，卷积核（5，5），步长（2，2），激活函数LeakyReLU()，丢失比例0.3
平铺层：
全连接层：
'''

def make_discriminator_model():
    model=tf.keras.Sequential()
    model.add(layers.Conv2D(64,(5,5),(2,2), padding='same', input_shape=[28, 28, 1]))
    model.add(layers.LeakyReLU())
    model.add(layers.Dropout(0.3))

    model.add(layers.Conv2D(128,(5,5),(2,2),padding='same'))
    model.add(layers.LeakyReLU())
    model.add(layers.Dropout(0.3))

    model.add(layers.Flatten())

    model.add(layers.Dense(1))

    return model


#定义损失函数和优化器
# 该方法返回计算交叉熵损失的辅助函数
cross_entropy = tf.keras.losses.BinaryCrossentropy(from_logits=True)

'''
判别器损失

判别器对真实图片的预测值与值全为 1 的数组进行对比，将判别器对伪造（生成的）图片的预测值与值全为 0 的数组进行对比。
判别器损失=真的交叉熵损失+假的交叉熵损失
'''
def discriminator_loss(real_output, fake_output):
    real_loss=cross_entropy(tf.ones_like(real_output),real_output)
    fake_loss=cross_entropy(tf.zeros_like(fake_output),fake_output)
    total_loss=real_loss+fake_loss

    return total_loss


'''
生成器损失
在生成图片上的判断结果与一个值全为 1 的数组进行对比
'''

def generator_loss(fake_output):
    return cross_entropy(tf.ones_like(fake_output),fake_output)


'''
优化器
'''
generator_optimizer=tf.keras.optimizers.Adam(0.0001)
discriminator_optimizer=tf.keras.optimizers.Adam(0.0001)

#定义训练循环

EPOCHS = 50
noise_dim = 100
num_examples_to_generate = 16
# BATCH_SIZE=256

# 我们将重复使用该种子
seed = tf.random.normal([num_examples_to_generate, noise_dim])

'''
生成噪声
由生成器生成图片
判别器生成真实图片特征，生成图片特征
计算损失值
设置循环梯度
设置循环优化器
'''

generator = make_generator_model()
discriminator = make_discriminator_model()

@tf.function
def train_step(images):
    noise = tf.random.normal([BATCH_SIZE, noise_dim])

    with tf.GradientTape() as gen_tape, tf.GradientTape() as disc_tape:
        gen_images=generator(noise, training=True)
        real_output=discriminator(images,training=True)
        fake_output=discriminator(gen_images,training=True)

        gen_loss=generator_loss(fake_output)
        disc_loss=discriminator_loss(real_output, fake_output)

    gradients_of_generator=gen_tape.gradient(gen_loss,generator.trainable_variables)
    gradients_of_discriminator= disc_tape.gradient(disc_loss,discriminator.trainable_variables)

    generator_optimizer.apply_gradients(zip(gradients_of_generator,generator.trainable_variables))
    discriminator_optimizer.apply_gradients(zip(gradients_of_discriminator,discriminator.trainable_variables))


checkpoint_dir = './training_checkpoints'
checkpoint_prefix = os.path.join(checkpoint_dir, "ckpt")
checkpoint = tf.train.Checkpoint(generator_optimizer=generator_optimizer,
                                 discriminator_optimizer=discriminator_optimizer,
                                 generator=generator,
                                 discriminator=discriminator)


#训练模型

def train(dataset, epochs):
    for epoch in range(epochs):
        start = time.time()

        for image_batch in dataset:
            train_step(image_batch)
            if epoch % 10 ==0:
                generate_and_save_images(generator,epochs,seed)
                checkpoint.save(file_prefix=checkpoint_prefix)
            print('Time for epoch {} is {} sec'.format(epoch + 1, time.time() - start))


def generate_and_save_images(model, epoch, test_input):
  # 注意 training` 设定为 False
  # 因此，所有层都在推理模式下运行（batchnorm）。
  predictions = model(test_input, training=False)

  fig = plt.figure(figsize=(4,4))

  for i in range(predictions.shape[0]):
      # plt.subplot(4, 4, i+1)
      plt.imshow(predictions[i, :, :, 0] * 127.5 + 127.5, cmap='gray')
      plt.axis('off')

  # plt.savefig('image_at_epoch_{:04d}.png'.format(epoch))
  plt.show()


(train_images, train_labels), (_, _) = tf.keras.datasets.mnist.load_data()
train_images = train_images.reshape(train_images.shape[0], 28, 28, 1).astype('float32')
train_images = (train_images - 127.5) / 127.5 # 将图片标准化到 [-1, 1] 区间内
BUFFER_SIZE = 60000
BATCH_SIZE = 256
# 批量化和打乱数据
train_dataset = tf.data.Dataset.from_tensor_slices(train_images).shuffle(BUFFER_SIZE).batch(BATCH_SIZE)


train(train_dataset, EPOCHS)
checkpoint.restore(tf.train.latest_checkpoint(checkpoint_dir))