mirror of
https://github.com/Estom/notes.git
synced 2026-04-05 11:57:37 +08:00
389 lines
13 KiB
Markdown
389 lines
13 KiB
Markdown
# 深度卷积生成对抗网络
|
||
|
||
> 原文:[https://tensorflow.google.cn/tutorials/generative/dcgan](https://tensorflow.google.cn/tutorials/generative/dcgan)
|
||
|
||
**Note:** 我们的 TensorFlow 社区翻译了这些文档。因为社区翻译是尽力而为, 所以无法保证它们是最准确的,并且反映了最新的 [官方英文文档](https://tensorflow.google.cn/?hl=en)。如果您有改进此翻译的建议, 请提交 pull request 到 [tensorflow/docs](https://github.com/tensorflow/docs) GitHub 仓库。要志愿地撰写或者审核译文,请加入 [docs-zh-cn@tensorflow.org Google Group](https://groups.google.com/a/tensorflow.org/forum/#!forum/docs-zh-cn)。
|
||
|
||
本教程演示了如何使用[深度卷积生成对抗网络](https://arxiv.org/pdf/1511.06434.pdf)(DCGAN)生成手写数字图片。该代码是使用 [Keras Sequential API](https://tensorflow.google.cn/guide/keras) 与 [`tf.GradientTape`](https://tensorflow.google.cn/api_docs/python/tf/GradientTape) 训练循环编写的。
|
||
|
||
## 什么是生成对抗网络?
|
||
|
||
[生成对抗网络](https://arxiv.org/abs/1406.2661)(GANs)是当今计算机科学领域最有趣的想法之一。两个模型通过对抗过程同时训练。一个*生成器*(“艺术家”)学习创造看起来真实的图像,而*判别器*(“艺术评论家”)学习区分真假图像。
|
||
|
||

|
||
|
||
训练过程中,*生成器*在生成逼真图像方面逐渐变强,而*判别器*在辨别这些图像的能力上逐渐变强。当*判别器*不再能够区分真实图片和伪造图片时,训练过程达到平衡。
|
||
|
||

|
||
|
||
本笔记在 MNIST 数据集上演示了该过程。下方动画展示了当训练了 50 个 epoch (全部数据集迭代 50 次) 时*生成器*所生成的一系列图片。图片从随机噪声开始,随着时间的推移越来越像手写数字。
|
||
|
||

|
||
|
||
要了解关于 GANs 的更多信息,我们建议参阅 MIT 的 [深度学习入门](http://introtodeeplearning.com/) 课程。
|
||
|
||
### Import TensorFlow and other libraries
|
||
|
||
```py
|
||
import tensorflow as tf
|
||
```
|
||
|
||
```py
|
||
tf.__version__
|
||
```
|
||
|
||
```py
|
||
'2.3.0'
|
||
|
||
```
|
||
|
||
```py
|
||
# 用于生成 GIF 图片
|
||
pip install -q imageio
|
||
|
||
```
|
||
|
||
```py
|
||
WARNING: You are using pip version 20.2.2; however, version 20.2.3 is available.
|
||
You should consider upgrading via the '/tmpfs/src/tf_docs_env/bin/python -m pip install --upgrade pip' command.
|
||
|
||
```
|
||
|
||
```py
|
||
import glob
|
||
import imageio
|
||
import matplotlib.pyplot as plt
|
||
import numpy as np
|
||
import os
|
||
import PIL
|
||
from tensorflow.keras import layers
|
||
import time
|
||
|
||
from IPython import display
|
||
```
|
||
|
||
### 加载和准备数据集
|
||
|
||
您将使用 MNIST 数据集来训练生成器和判别器。生成器将生成类似于 MNIST 数据集的手写数字。
|
||
|
||
```py
|
||
(train_images, train_labels), (_, _) = tf.keras.datasets.mnist.load_data()
|
||
```
|
||
|
||
```py
|
||
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
|
||
```
|
||
|
||
```py
|
||
BUFFER_SIZE = 60000
|
||
BATCH_SIZE = 256
|
||
```
|
||
|
||
```py
|
||
# 批量化和打乱数据
|
||
train_dataset = tf.data.Dataset.from_tensor_slices(train_images).shuffle(BUFFER_SIZE).batch(BATCH_SIZE)
|
||
```
|
||
|
||
## 创建模型
|
||
|
||
生成器和判别器均使用 [Keras Sequential API](https://tensorflow.google.cn/guide/keras#sequential_model) 定义。
|
||
|
||
### 生成器
|
||
|
||
生成器使用 [`tf.keras.layers.Conv2DTranspose`](https://tensorflow.google.cn/api_docs/python/tf/keras/layers/Conv2DTranspose) (上采样)层来从种子(随机噪声)中产生图片。以一个使用该种子作为输入的 `Dense` 层开始,然后多次上采样直到达到所期望的 28x28x1 的图片尺寸。注意除了输出层使用 tanh 之外,其他每层均使用 [`tf.keras.layers.LeakyReLU`](https://tensorflow.google.cn/api_docs/python/tf/keras/layers/LeakyReLU) 作为激活函数。
|
||
|
||
```py
|
||
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)))
|
||
assert model.output_shape == (None, 7, 7, 256) # 注意:batch size 没有限制
|
||
|
||
model.add(layers.Conv2DTranspose(128, (5, 5), strides=(1, 1), padding='same', use_bias=False))
|
||
assert model.output_shape == (None, 7, 7, 128)
|
||
model.add(layers.BatchNormalization())
|
||
model.add(layers.LeakyReLU())
|
||
|
||
model.add(layers.Conv2DTranspose(64, (5, 5), strides=(2, 2), padding='same', use_bias=False))
|
||
assert model.output_shape == (None, 14, 14, 64)
|
||
model.add(layers.BatchNormalization())
|
||
model.add(layers.LeakyReLU())
|
||
|
||
model.add(layers.Conv2DTranspose(1, (5, 5), strides=(2, 2), padding='same', use_bias=False, activation='tanh'))
|
||
assert model.output_shape == (None, 28, 28, 1)
|
||
|
||
return model
|
||
```
|
||
|
||
使用(尚未训练的)生成器创建一张图片。
|
||
|
||
```py
|
||
generator = make_generator_model()
|
||
|
||
noise = tf.random.normal([1, 100])
|
||
generated_image = generator(noise, training=False)
|
||
|
||
plt.imshow(generated_image[0, :, :, 0], cmap='gray')
|
||
```
|
||
|
||
```py
|
||
<matplotlib.image.AxesImage at 0x7f01d26074a8>
|
||
|
||
```
|
||
|
||

|
||
|
||
### 判别器
|
||
|
||
判别器是一个基于 CNN 的图片分类器。
|
||
|
||
```py
|
||
def make_discriminator_model():
|
||
model = tf.keras.Sequential()
|
||
model.add(layers.Conv2D(64, (5, 5), strides=(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), strides=(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
|
||
```
|
||
|
||
使用(尚未训练的)判别器来对图片的真伪进行判断。模型将被训练为为真实图片输出正值,为伪造图片输出负值。
|
||
|
||
```py
|
||
discriminator = make_discriminator_model()
|
||
decision = discriminator(generated_image)
|
||
print (decision)
|
||
```
|
||
|
||
```py
|
||
tf.Tensor([[-0.00427552]], shape=(1, 1), dtype=float32)
|
||
|
||
```
|
||
|
||
## 定义损失函数和优化器
|
||
|
||
为两个模型定义损失函数和优化器。
|
||
|
||
```py
|
||
# 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
|
||
|
||
该方法量化判别器从判断真伪图片的能力。它将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。
|
||
|
||
```py
|
||
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
|
||
|
||
生成器损失量化其欺骗判别器的能力。直观来讲,如果生成器表现良好,判别器将会把伪造图片判断为真实图片(或 1)。这里我们将把判别器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。
|
||
|
||
```py
|
||
def generator_loss(fake_output):
|
||
return cross_entropy(tf.ones_like(fake_output), fake_output)
|
||
```
|
||
|
||
由于我们需要分别训练两个网络,判别器和生成器的优化器是不同的。
|
||
|
||
```py
|
||
generator_optimizer = tf.keras.optimizers.Adam(1e-4)
|
||
discriminator_optimizer = tf.keras.optimizers.Adam(1e-4)
|
||
```
|
||
|
||
### 保存检查点
|
||
|
||
本笔记还演示了如何保存和恢复模型,这在长时间训练任务被中断的情况下比较有帮助。
|
||
|
||
```py
|
||
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)
|
||
```
|
||
|
||
## 定义训练循环
|
||
|
||
```py
|
||
EPOCHS = 50
|
||
noise_dim = 100
|
||
num_examples_to_generate = 16
|
||
|
||
# 我们将重复使用该种子(因此在动画 GIF 中更容易可视化进度)
|
||
seed = tf.random.normal([num_examples_to_generate, noise_dim])
|
||
```
|
||
|
||
训练循环在生成器接收到一个随机种子作为输入时开始。该种子用于生产一张图片。判别器随后被用于区分真实图片(选自训练集)和伪造图片(由生成器生成)。针对这里的每一个模型都计算损失函数,并且计算梯度用于更新生成器与判别器。
|
||
|
||
```py
|
||
# 注意 `tf.function` 的使用
|
||
# 该注解使函数被“编译”
|
||
@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:
|
||
generated_images = generator(noise, training=True)
|
||
|
||
real_output = discriminator(images, training=True)
|
||
fake_output = discriminator(generated_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))
|
||
```
|
||
|
||
```py
|
||
def train(dataset, epochs):
|
||
for epoch in range(epochs):
|
||
start = time.time()
|
||
|
||
for image_batch in dataset:
|
||
train_step(image_batch)
|
||
|
||
# 继续进行时为 GIF 生成图像
|
||
display.clear_output(wait=True)
|
||
generate_and_save_images(generator,
|
||
epoch + 1,
|
||
seed)
|
||
|
||
# 每 15 个 epoch 保存一次模型
|
||
if (epoch + 1) % 15 == 0:
|
||
checkpoint.save(file_prefix = checkpoint_prefix)
|
||
|
||
print ('Time for epoch {} is {} sec'.format(epoch + 1, time.time()-start))
|
||
|
||
# 最后一个 epoch 结束后生成图片
|
||
display.clear_output(wait=True)
|
||
generate_and_save_images(generator,
|
||
epochs,
|
||
seed)
|
||
```
|
||
|
||
**生成与保存图片**
|
||
|
||
```py
|
||
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()
|
||
```
|
||
|
||
## 训练模型
|
||
|
||
调用上面定义的 `train()` 方法来同时训练生成器和判别器。注意,训练 GANs 可能是棘手的。重要的是,生成器和判别器不能够互相压制对方(例如,他们以相似的学习率训练)。
|
||
|
||
在训练之初,生成的图片看起来像是随机噪声。随着训练过程的进行,生成的数字将越来越真实。在大概 50 个 epoch 之后,这些图片看起来像是 MNIST 数字。使用 Colab 中的默认设置可能需要大约 1 分钟每 epoch。
|
||
|
||
```py
|
||
%%time
|
||
train(train_dataset, EPOCHS)
|
||
```
|
||
|
||

|
||
|
||
```py
|
||
CPU times: user 1min 52s, sys: 11.7 s, total: 2min 4s
|
||
Wall time: 3min 22s
|
||
|
||
```
|
||
|
||
恢复最新的检查点。
|
||
|
||
```py
|
||
checkpoint.restore(tf.train.latest_checkpoint(checkpoint_dir))
|
||
```
|
||
|
||
```py
|
||
<tensorflow.python.training.tracking.util.CheckpointLoadStatus at 0x7f0118537668>
|
||
|
||
```
|
||
|
||
## 创建 GIF
|
||
|
||
```py
|
||
# 使用 epoch 数生成单张图片
|
||
def display_image(epoch_no):
|
||
return PIL.Image.open('image_at_epoch_{:04d}.png'.format(epoch_no))
|
||
```
|
||
|
||
```py
|
||
display_image(EPOCHS)
|
||
```
|
||
|
||

|
||
|
||
使用训练过程中生成的图片通过 `imageio` 生成动态 gif
|
||
|
||
```py
|
||
anim_file = 'dcgan.gif'
|
||
|
||
with imageio.get_writer(anim_file, mode='I') as writer:
|
||
filenames = glob.glob('image*.png')
|
||
filenames = sorted(filenames)
|
||
last = -1
|
||
for i,filename in enumerate(filenames):
|
||
frame = 2*(i**0.5)
|
||
if round(frame) > round(last):
|
||
last = frame
|
||
else:
|
||
continue
|
||
image = imageio.imread(filename)
|
||
writer.append_data(image)
|
||
image = imageio.imread(filename)
|
||
writer.append_data(image)
|
||
|
||
import IPython
|
||
if IPython.version_info > (6,2,0,''):
|
||
display.Image(filename=anim_file)
|
||
```
|
||
|
||
如果您正在使用 Colab,您可以通过如下代码下载动画:
|
||
|
||
```py
|
||
try:
|
||
from google.colab import files
|
||
except ImportError:
|
||
pass
|
||
else:
|
||
files.download(anim_file)
|
||
```
|
||
|
||
## 下一步
|
||
|
||
本教程展示了实现和训练 GAN 模型所需的全部必要代码。接下来,您可能想尝试其他数据集,例如大规模名人面部属性(CelebA)数据集 [在 Kaggle 上获取](https://www.kaggle.com/jessicali9530/celeba-dataset)。要了解更多关于 GANs 的信息,我们推荐参阅 [NIPS 2016 教程: 生成对抗网络](https://arxiv.org/abs/1701.00160)。 |