DCGAN：生成动漫头像

个人博客：http://www.chenjianqu.com/

原文链接：http://www.chenjianqu.com/show-55.html

上一篇博文《生成对抗网络(GAN)原理和实现》介绍了GAN的原理和实现了GAN最原始的算法，今天这篇博客写一下读DCGAN论文所做的笔记，并基于网上的tensorflow开源代码实现DCGANs的动漫头像的生成。

论文笔记

论文：《UNSUPERVISED REPRESENTATION LEARNING WITH DEEP CONVOLUTIONAL GENERATIVE ADVERSARIAL NETWORKS》

1．解决什么问题

CNN在无监督学习上的应用。

Convolutinal GANs训练不稳定。

可视化GAN。

2．使用什么方法

提出了一系列技术去稳定的训练Convolutional GAN，叫做DCGAN。

使用训练好的判别器去做图像分类任务。

使用guided backpropagation可视化GAN。

证明生成器具有可向量运算的特性。

3．实验效果如何

DCGAN的G和D都能学习到分布到分布的映射的层级表示.

使用GAN作为特征抽取器，DCGAN在imagenet-1k上训练，将特征层用于SVM在CIFAR-10上分类，达到82.8%的准确率，超过了基于K-means的基线。

向量表示实验和可视化效果还不错。

4．仍存在的问题

模型仍有一些不稳定性存在，长时间训练的时候有时会有时会将滤波器的一个子集折叠成一个单一的振荡模式。

论文细节：

让DCGANs稳定训练的方法：

1.在判别器使用stride=2的卷积层替代池化层，在生成器使用frational-stride(转置卷积)替代池化层。

2.在判别器和生成器使用batchnorm

3.去掉全连接隐藏层

4.在生成器，除了输出层使用tanh激活函数，其他层使用relu激活函数。

5.在判别器，所有层都使用LeakyRelu激活函数。

使用到的数据集：

Large-scale Scene Understanding (LSUN)

Imagenet-1k

他们自己在网络上爬下来的人脸数据集

训练细节：

将图片的数据方位转换至[-1,1]，使用SGD的batch_size=128，权重采用高斯分布的随机初始化，均值为0，方差为0.02。LeakyReLU的leak的斜率为0.2。使用Adam优化器，学习率为0.0002,beta1参数为0.9

基于LSUN数据集的生成网络如下：

代码实现

数据集

动漫妹子的数据集爬取的代码网上有很多，爬下来后使用opencv的级联分类器可以截取头像。当然我这里使用的网上找的现成的数据集，我把数据集放在百度网盘：链接：https://pan.baidu.com/s/1MjzVS0RZ8jHkOspgX2yFNw 提取码：nj7l 。

2.定义DCGAN的网络，并封装成类。

网络如下：

其中两个判别器的变量是共享的。

生成器的网络如下：

判别器的网络如下：

代码如下：

import math
import tensorflow as tf
slim = tf.contrib.slim
class DCGAN(object):def __init__(self, is_training,#是否训练generator_depth=64,#生成器最后的反卷积层的输出深度discriminator_depth=64,#判别器的输入深度final_size=32,#生成器输出图片的大小num_outputs=3,#生成器输出图片的通道fused_batch_norm=False#batchnorm的fuse是否为true):self._is_training = is_trainingself._generator_depth = generator_depthself._discirminator_depth = discriminator_depthself._final_size = final_sizeself._num_outputs = num_outputsself._fused_batch_norm = fused_batch_norm#构造判别器     def discriminator(self, inputs,#输入batch图片depth=64,# Number of channels in first convolution layer.is_training=True,reuse=None,#是否参数重用scope='Discriminator',#scopefused_batch_norm=False ):normalizer_fn = slim.batch_normnormalizer_fn_args = {'is_training': is_training,'zero_debias_moving_mean': True,'fused': fused_batch_norm}height = inputs.get_shape().as_list()[1]end_points = {}with tf.variable_scope(scope, values=[inputs], reuse=reuse) as scope:with slim.arg_scope([normalizer_fn], **normalizer_fn_args):#判别器使用stride=2 kernel_size=4的cnn下采样，激活函数默认为leaky_reluwith slim.arg_scope([slim.conv2d], stride=2, kernel_size=4,activation_fn=tf.nn.leaky_relu):net = inputsfor i in range(int(math.log(height, 2))): #一共下采样log2(height)次scope = 'conv%i' % (i+1)current_depth = depth * 2**i #每次深度翻倍normalizer_fn_ = None if i == 0 else normalizer_fnnet = slim.conv2d(net, num_outputs=current_depth, normalizer_fn=normalizer_fn_,scope=scope)end_points[scope] = net #保存tensor至词典#使用1x1的卷积代替全连接层输出logits = slim.conv2d(net, 1, kernel_size=1, stride=1,padding='VALID', normalizer_fn=None,activation_fn=None)logits = tf.reshape(logits, [-1, 1])end_points['logits'] = logits #保存tensor至词典return logits, end_points#构造生成器def generator(self,inputs,#输入向量depth=64,#转置卷积最后的通道数final_size=32,#生成的图片大小num_outputs=3,#输出的图片通道is_training=True,#是否正在训练reuse=None,#是否参数重用scope='Generator',#生成器的变量域fused_batch_norm=False#If 'True', use a faster, fused implementationof batch normalization.):normalizer_fn = slim.batch_normnormalizer_fn_args = { 'is_training': is_training,'zero_debias_moving_mean': True, 'fused': fused_batch_norm}#检查final_size必须是2的幂次if math.log(final_size, 2) != int(math.log(final_size, 2)):raise ValueError("'final_size' (%i) must be a power of 2." % final_size)#final_size大于8if final_size < 8: raise ValueError("'final_size' (%i) must be greater than 8."% final_size)end_points = {}num_layers = int(math.log(final_size, 2)) - 1 #转置卷积的层数，是log2(64)-1=5with tf.variable_scope(scope, values=[inputs], reuse=reuse) as scope:with slim.arg_scope([normalizer_fn], **normalizer_fn_args):#normalizer共享参数域with slim.arg_scope([slim.conv2d_transpose],normalizer_fn=normalizer_fn,stride=2, kernel_size=4):#转置卷积共享参数域net = tf.expand_dims(tf.expand_dims(inputs, 1), 1)#将输入[batch,n]拓展为[batch,n,1,1]current_depth = depth * 2 ** (num_layers - 1) #通道数scope = 'deconv1'net = slim.conv2d_transpose(net, current_depth, stride=1, padding='VALID', scope=scope)#第一次转置卷积的stride=1end_points[scope] = net #保存tensor#多层转置卷积for i in range(2, num_layers):scope = 'deconv%i' % icurrent_depth = depth * 2 * (num_layers - i)net = slim.conv2d_transpose(net, current_depth, scope=scope)end_points[scope] = net#最后的反卷积层scope = 'deconv%i' % num_layersnet = slim.conv2d_transpose(net, depth, normalizer_fn=None,activation_fn=None, scope=scope)end_points[scope] = net#卷积得到图片scope = 'logits'logits = slim.conv2d(net,num_outputs,normalizer_fn=None,activation_fn=tf.nn.tanh,kernel_size=1,stride=1,padding='VALID',scope=scope)end_points[scope] = logitslogits.get_shape().assert_has_rank(4)logits.get_shape().assert_is_compatible_with([None, final_size, final_size, num_outputs])return logits, end_points#构造dcgandef dcgan_model(self, real_data,#输入batch图片generator_inputs,#输入向量generator_scope='Generator',#生成器变量域 当你想重用已经训练的网络权重时discirminator_scope='Discriminator',#判别器变量域check_shapes=True):#检查生成器生成的张量形状#创建生成器with tf.variable_scope(generator_scope) as gen_scope:generated_data, _ = self.generator(generator_inputs, self._generator_depth, self._final_size,self._num_outputs, self._is_training)#创建判别器with tf.variable_scope(discirminator_scope) as dis_scope:discriminator_gen_outputs, _ = self.discriminator(generated_data, self._discirminator_depth, self._is_training)with tf.variable_scope(dis_scope, reuse=True):discriminator_real_outputs, _ = self.discriminator(real_data, self._discirminator_depth, self._is_training)#检查生成器生成的数据形状是否和真实数据形状一样if check_shapes:if not generated_data.shape.is_compatible_with(real_data.shape):raise ValueError('Generator output shape (%s) must be the shape as real data (%s).'% (generated_data.shape, real_data.shape))#获得训练的变量generator_variables = slim.get_trainable_variables(gen_scope)discriminator_variables = slim.get_trainable_variables(dis_scope)#返回tensorreturn {'generated_data': generated_data,'discriminator_gen_outputs': discriminator_gen_outputs,'discriminator_real_outputs': discriminator_real_outputs,'generator_variables': generator_variables,'discriminator_variables': discriminator_variables}#预测，生成图片def predict(self, generator_inputs):logits, _ = self.generator(generator_inputs, self._generator_depth, self._final_size, self._num_outputs,is_training=False)return logits#判别器的lossdef discriminator_loss(self, discriminator_real_outputs,discriminator_gen_outputs,label_smoothing=0.25): #使用label smoothing技术提高训练效果# -log((1 - label_smoothing) - sigmoid(D(x)))losses_on_real = slim.losses.sigmoid_cross_entropy(logits=discriminator_real_outputs,multi_class_labels=tf.ones_like(discriminator_real_outputs),label_smoothing=label_smoothing)loss_on_real = tf.reduce_mean(losses_on_real)# -log(- sigmoid(D(G(x))))losses_on_generated = slim.losses.sigmoid_cross_entropy(logits=discriminator_gen_outputs,multi_class_labels=tf.zeros_like(discriminator_gen_outputs))loss_on_generated = tf.reduce_mean(losses_on_generated)loss = loss_on_real + loss_on_generatedreturn {'dis_loss': loss,'dis_loss_on_real': loss_on_real, 'dis_loss_on_generated': loss_on_generated}#生成器的Lossdef generator_loss(self, discriminator_gen_outputs, label_smoothing=0.0):losses = slim.losses.sigmoid_cross_entropy(logits=discriminator_gen_outputs, multi_class_labels=tf.ones_like(discriminator_gen_outputs),label_smoothing=label_smoothing)loss = tf.reduce_mean(losses)return loss#计算dcgan的损失def loss(self, discriminator_real_outputs, discriminator_gen_outputs):with tf.name_scope('loss'):dis_loss_dict = self.discriminator_loss(discriminator_real_outputs,discriminator_gen_outputs)gen_loss = self.generator_loss(discriminator_gen_outputs)dis_loss_dict.update({'gen_loss': gen_loss})return dis_loss_dict

3.定义工具函数

import cv2
import glob
import numpy as np
import os
import tensorflow as tfDATA_DIR='D:/CV/datasets/cartoon_face/faces_not_oom'
IMAGE_SAVE_DIR='D:/Jupyter/GAN/gan_cartoon_image_data/save_pic'
LOG_DIR='D:/Jupyter/GAN/gan_cartoon_image_data/train_log'def read_images(image_files):images = []for image_path in image_files:image = cv2.imread(image_path)image = cv2.resize(image, (64, 64))image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)image = (image - 127.5) / 127.5 #将图片的取值范围设置为（-1,1）images.append(image)return np.array(images)
#产生图片的输入的随机向量
def get_next_batch(batch_size=8):"""Get a batch set of real images and random generated inputs."""images_path = os.path.join(DATA_DIR,'*.jpg')image_files_list = glob.glob(images_path) #匹配.jpg图片image_files_arr = np.array(image_files_list)selected_indices = np.random.choice(len(image_files_list), batch_size)#随机生成索引数组selected_image_files = image_files_arr[selected_indices]#选择一个批次的图片路径images = read_images(selected_image_files) #读取图片generated_inputs = np.random.uniform(low=-1, high=1.0, size=[batch_size, 64]) #随机产生64维的随机向量return images, generated_inputs
#保存图片
def write_images(generated_images, images_save_dir, num_step):#Scale images from [-1, 1] to [0, 255].generated_images = ((generated_images + 1) * 127.5).astype(np.uint8) #转换格式for j, image in enumerate(generated_images[:5]):image_name = 'generated_step{}_{}.jpg'.format(num_step+1, j+1)image_path = os.path.join(images_save_dir,image_name)image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)cv2.imwrite(image_path, image)

4.训练DCGAN

STEPS=30000#placeholder
real_data = tf.placeholder(tf.float32, shape=[None, 64, 64, 3], name='real_data')
generated_inputs = tf.placeholder(tf.float32, [None, 64], name='generated_inputs')#创建gan
dcgan_model = DCGAN(is_training=True, final_size=64)
outputs_dict = dcgan_model.dcgan_model(real_data, generated_inputs)
generated_data = outputs_dict['generated_data'] #生成器的输出张量
generated_data_ = tf.identity(generated_data, name='generated_data')#y = tf.identity(x)是一个op操作表示将x的值赋予y
discriminator_gen_outputs = outputs_dict['discriminator_gen_outputs'] #假的输入时判别器的输出张量
discriminator_real_outputs = outputs_dict['discriminator_real_outputs'] #真实输入时判别器的输出张量
generator_variables = outputs_dict['generator_variables'] #生成器可训练的张量
discriminator_variables = outputs_dict['discriminator_variables'] #判别器可训练的张量#获取dcgan的损失值
loss_dict = dcgan_model.loss(discriminator_real_outputs,discriminator_gen_outputs)
discriminator_loss = loss_dict['dis_loss']
discriminator_loss_on_real = loss_dict['dis_loss_on_real']
discriminator_loss_on_generated = loss_dict['dis_loss_on_generated']
generator_loss = loss_dict['gen_loss']#将变量的损失值写入Loss
tf.summary.scalar('discriminator_loss', discriminator_loss)
tf.summary.scalar('discriminator_loss_on_real', discriminator_loss_on_real)
tf.summary.scalar('discriminator_loss_on_generated',discriminator_loss_on_generated)
tf.summary.scalar('generator_loss', generator_loss)
merged_summary = tf.summary.merge_all(key=tf.GraphKeys.SUMMARIES)#创建判别器的优化器
discriminator_optimizer = tf.train.AdamOptimizer(learning_rate=0.0004,beta1=0.5)
discriminator_train_step = discriminator_optimizer.minimize(discriminator_loss, var_list=discriminator_variables)#创建生成器的优化器
generator_optimizer = tf.train.AdamOptimizer(learning_rate=0.0001,beta1=0.5)
generator_train_step = generator_optimizer.minimize(generator_loss, var_list=generator_variables)saver = tf.train.Saver(var_list=tf.global_variables())init = tf.global_variables_initializer()with tf.Session() as sess:sess.run(init)# 保存图writer = tf.summary.FileWriter(LOG_DIR, sess.graph)#获取图片和向量，用于测试效果fixed_images, fixed_generated_inputs = get_next_batch()#断点续训ckpt=tf.train.get_checkpoint_state(LOG_DIR)if(ckpt and ckpt.model_checkpoint_path):saver.restore(sess,ckpt.model_checkpoint_path)for i in range(STEPS):#更新判别器batch_images, batch_generated_inputs = get_next_batch()train_dict = {real_data: batch_images,generated_inputs: batch_generated_inputs}sess.run(discriminator_train_step, feed_dict=train_dict)#更新生成器3次batch_images, batch_generated_inputs = get_next_batch()train_dict = {real_data: batch_images,generated_inputs: batch_generated_inputs}sess.run(generator_train_step, feed_dict=train_dict)batch_images, batch_generated_inputs = get_next_batch()train_dict = {real_data: batch_images,generated_inputs: batch_generated_inputs}sess.run(generator_train_step, feed_dict=train_dict)batch_images, batch_generated_inputs = get_next_batch()train_dict = {real_data: batch_images,generated_inputs: batch_generated_inputs}sess.run(generator_train_step, feed_dict=train_dict)#200轮输出一次图片对比效果if (i+1) % 200 == 0:batch_images = fixed_imagesbatch_generated_inputs = fixed_generated_inputselse:batch_images, batch_generated_inputs = get_next_batch()train_dict = {real_data: batch_images,generated_inputs: batch_generated_inputs}#获得summary和生成的图片summary, generated_images  = sess.run([merged_summary, generated_data], feed_dict=train_dict)#写summarywriter.add_summary(summary, i+1)if(i%10==0):print(i,end=' ')if (i+1) % 200 == 0:#模型model_save_path = os.path.join(LOG_DIR, 'model.ckpt')saver.save(sess, save_path=model_save_path, global_step=i+1)#保存图片write_images(generated_images, IMAGE_SAVE_DIR, i)writer.close()

由于多次OOM，因此loss曲线太乱了，就不放出来了。这是训练不知道多少轮后的效果：

参考文献

[1]Alec Radford & Luke Metz,Soumith Chintala.UNSUPERVISED REPRESENTATION LEARNING WITH DEEP CONVOLUTIONAL GENERATIVE ADVERSARIAL NETWORKS.2016-1-06

[2]公输睚信.TensorFlow 从零开始实现深度卷积生成对抗网络（DCGAN）.https://www.jianshu.com/p/4d8473070ae7 .2018-05-27

DCGAN：生成动漫头像相关推荐

vs2019 利用Pytorch和TensorFlow分别实现DCGAN生成动漫头像
这是针对于博客vs2019安装和使用教程(详细)的DCGAN生成动漫头像项目新建示例目录一.DCGAN架构及原理二.项目结构 1.TensorFlow 2.Pytorch 三.数据集下载(两种方 ...
DCGAN生成动漫头像【学习】
DCGAN生成动漫头像在假期看了李宏毅老师的GAN的介绍,看到了课后题DCGAN生成动漫头像的作业,实现一下.记录学习过程. 参考的文章: [Keras] 基于GAN自动生成动漫头像因为使用的是t ...
pytorch：DCGAN生成动漫头像
动漫头像数据集下载地址:动漫头像数据集_百度云连接,DCGAN论文下载地址: https://arxiv.org/abs/1511.06434 数据集里面的图片是这个样子的: 这是DCGAN的主要改进 ...
使用TensorFlow2.0搭建DCGAN生成动漫头像（内含生成过程GIF图）
文章目录生成对抗网络介绍一.造假二.训练判别器三.训练生成器 DCGAN介绍搭建DCGAN 数据来源必要工作读取数据构建生成器构建判别器连接模型连接图片生成函数训练生成对抗 ...
通过PyTorch用DCGAN生成动漫头像
数据集数据集我们用AnimeFaces数据集,共5万多张动漫头像. 链接:https://pan.baidu.com/s/1cp-A8ZV74YBelkSuKxuM6A 提取码:face 要把所有的 ...
基于Tensorflow和DCGAN生成动漫头像实践（二）
本篇内容为动漫头像生成的主要代码部分,第一次写这种代码,从读取数据到生成走了一个完整的流程.创建TFrecord过程可以看上一篇内容. 代码内容: #!/usr/bin/env python2 # - ...
DCGAN生成动漫头像（附代码）
DCGAN.顾名思义,就是深度卷积生成对抗神经网络,也就是引入了卷积的,但是它用的是反卷积,就是卷积的反操作. 我们看看DCGAN的图: 生成器开始输入的是噪声数据,然后经过一个全连接层,再把全连接层 ...
基于DCGAN的动漫头像生成
基于DCGAN的动漫头像生成数据数据集:动漫图库爬虫获得,经过数据清洗,裁剪得到动漫头像.分辨率为3 * 96 * 96,共5万多张动漫头像的图片,从知乎用户何之源处下载. 生成器:输入为随机噪声 ...
有趣的图像生成——使用DCGAN与pytorch生成动漫头像
有趣的图像生成--使用DCGAN与pytorch生成动漫头像文章目录有趣的图像生成--使用DCGAN与pytorch生成动漫头像一.源码下载二.什么是DCGAN 三.DCGAN的实现 1.** ...
pytorch实现DCGAN生成动漫人物头像
pytorch实现DCGAN生成动漫人物头像 DCGAN原理参考这一系列文章数据集 21551张64*64动漫人物头像生成效果训练1个epoch(emm-) 训练10个epoch(起码有颜色了 ...

DCGAN：生成动漫头像

DCGAN：生成动漫头像相关推荐

最新文章

热门文章