代码实现参考：https://github.com/affinelayer/pix2pix-tensorflow，论文参考《Image-to-Image Translation with Conditional Adversarial Nets》，以下实现只在facades数据集上训练，运行代码前需要先下载该数据集。地带了2000次epoch，下过如下（左边为输入图像，中间为生成的图像，右边为真实图像）：
#coding=utf-8
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import tensorflow as tf
import numpy as np
import os
import glob
import random
import collections
import math
import time
# https://github.com/affinelayer/pix2pix-tensorflow
train_input_dir="E:/testdata/facades/train/" # 训练集输入
train_output_dir="E:/testdata/facades/train_out/" # 训练集输出
test_input_dir="E:/testdata/facades/val/" # 测试集输入
test_output_dir="E:/testdata/facades/test_out/" # 测试集的输出
checkpoint="E:/testdata/facades/train_out/" # 保存结果的目录
seed=None
max_steps=None # number of training steps (0 to disable)
max_epochs=200 # number of training epochs
progress_freq=50 # display progress every progress_freq steps
trace_freq=0 # trace execution every trace_freq steps
display_freq=50 # write current training images every display_freq steps
save_freq=500 # save model every save_freq steps, 0 to disable
separable_conv=False # use separable convolutions in the generator
aspect_ratio=1.0 # aspect ratio of output images (width/height)
batch_size=1 # help="number of images in batch")
which_direction="BtoA" # choices=["AtoB", "BtoA"])
ngf=64 # help="number of generator filters in first conv layer")
ndf=64 # help="number of discriminator filters in first conv layer")
scale_size=286 # help="scale images to this size before cropping to 256x256")
flip=True # flip images horizontally
no_flip=True # don't flip images horizontally
lr=0.0002 # initial learning rate for adam
beta1=0.5 # momentum term of adam
l1_weight=100.0 # weight on L1 term for generator gradient
gan_weight=1.0 # weight on GAN term for generator gradient
output_filetype="png" # 输出图像的格式
EPS = 1e-12 # 极小数，防止梯度为损失为0
CROP_SIZE = 256 # 图片的裁剪大小
# 命名元组,用于存放加载的数据集合创建好的模型
Examples = collections.namedtuple("Examples", "paths, inputs, targets, count, steps_per_epoch")
Model = collections.namedtuple("Model", "outputs, predict_real, predict_fake, discrim_loss, discrim_grads_and_vars, gen_loss_GAN, gen_loss_L1, gen_grads_and_vars, train")
# 图像预处理 [0, 1] => [-1, 1]
def preprocess(image):
with tf.name_scope("preprocess"):
return image * 2 - 1
# 图像后处理[-1, 1] => [0, 1]
def deprocess(image):
with tf.name_scope("deprocess"):
return (image + 1) / 2
# 判别器的卷积定义，batch_input为 [ batch , 256 , 256 , 6 ]
def discrim_conv(batch_input, out_channels, stride):
# [ batch , 256 , 256 , 6 ] ===>[ batch , 258 , 258 , 6 ]
padded_input = tf.pad(batch_input, [[0, 0], [1, 1], [1, 1], [0, 0]], mode="CONSTANT")
'''
[0,0]: 第一维batch大小不扩充
[1,1]：第二维图像宽度左右各扩充一列，用0填充
[1,1]：第三维图像高度上下各扩充一列，用0填充
[0,0]：第四维图像通道不做扩充
'''
return tf.layers.conv2d(padded_input, out_channels, kernel_size=4, strides=(stride, stride), padding="valid", kernel_initializer=tf.random_normal_initializer(0, 0.02))
# 生成器的卷积定义，卷积核为4*4，步长为2，输出图像为输入的一半
def gen_conv(batch_input, out_channels):
# [batch, in_height, in_width, in_channels] => [batch, out_height, out_width, out_channels]
initializer = tf.random_normal_initializer(0, 0.02)
if separable_conv:
return tf.layers.separable_conv2d(batch_input, out_channels, kernel_size=4, strides=(2, 2), padding="same", depthwise_initializer=initializer, pointwise_initializer=initializer)
else:
return tf.layers.conv2d(batch_input, out_channels, kernel_size=4, strides=(2, 2), padding="same", kernel_initializer=initializer)
# 生成器的反卷积定义
def gen_deconv(batch_input, out_channels):
# [batch, in_height, in_width, in_channels] => [batch, out_height, out_width, out_channels]
initializer = tf.random_normal_initializer(0, 0.02)
if separable_conv:
_b, h, w, _c = batch_input.shape
resized_input = tf.image.resize_images(batch_input, [h * 2, w * 2], method=tf.image.ResizeMethod.NEAREST_NEIGHBOR)
return tf.layers.separable_conv2d(resized_input, out_channels, kernel_size=4, strides=(1, 1), padding="same", depthwise_initializer=initializer, pointwise_initializer=initializer)
else:
return tf.layers.conv2d_transpose(batch_input, out_channels, kernel_size=4, strides=(2, 2), padding="same", kernel_initializer=initializer)
# 定义LReLu激活函数
def lrelu(x, a):
with tf.name_scope("lrelu"):
# adding these together creates the leak part and linear part
# then cancels them out by subtracting/adding an absolute value term
# leak: a*x/2 - a*abs(x)/2
# linear: x/2 + abs(x)/2
# this block looks like it has 2 inputs on the graph unless we do this
x = tf.identity(x)
return (0.5 * (1 + a)) * x + (0.5 * (1 - a)) * tf.abs(x)
# 批量归一化图像
def batchnorm(inputs):
return tf.layers.batch_normalization(inputs, axis=3, epsilon=1e-5, momentum=0.1, training=True, gamma_initializer=tf.random_normal_initializer(1.0, 0.02))
# 检查图像的维度
def check_image(image):
assertion = tf.assert_equal(tf.shape(image)[-1], 3, message="image must have 3 color channels")
with tf.control_dependencies([assertion]):
image = tf.identity(image)
if image.get_shape().ndims not in (3, 4):
raise ValueError("image must be either 3 or 4 dimensions")
# make the last dimension 3 so that you can unstack the colors
shape = list(image.get_shape())
shape[-1] = 3
image.set_shape(shape)
return image
# 去除文件的后缀，获取文件名
def get_name(path):
# os.path.basename(),返回path最后的文件名。若path以/或\结尾，那么就会返回空值。
# os.path.splitext(),分离文件名与扩展名；默认返回(fname,fextension)元组
name, _ = os.path.splitext(os.path.basename(path))
return name
# 加载数据集，从文件读取-->解码-->归一化--->拆分为输入和目标-->像素转为[-1,1]-->转变形状
def load_examples(input_dir):
if input_dir is None or not os.path.exists(input_dir):
raise Exception("input_dir does not exist")
# 匹配第一个参数的路径中所有的符合条件的文件，并将其以list的形式返回。
input_paths = glob.glob(os.path.join(input_dir, "*.jpg"))
# 图像解码器
decode = tf.image.decode_jpeg
if len(input_paths) == 0:
input_paths = glob.glob(os.path.join(input_dir, "*.png"))
decode = tf.image.decode_png
if len(input_paths) == 0:
raise Exception("input_dir contains no image files")
# 如果文件名是数字，则用数字进行排序，否则用字母排序
if all(get_name(path).isdigit() for path in input_paths):
input_paths = sorted(input_paths, key=lambda path: int(get_name(path)))
else:
input_paths = sorted(input_paths)
sess = tf.Session()
with tf.name_scope("load_images"):
# 把我们需要的全部文件打包为一个tf内部的queue类型，之后tf开文件就从这个queue中取目录了，
# 如果是训练模式时，shuffle为True
path_queue = tf.train.string_input_producer(input_paths, shuffle=True)
# Read的输出将是一个文件名（key）和该文件的内容（value,每次读取一个文件，分多次读取）。
reader = tf.WholeFileReader()
paths, contents = reader.read(path_queue)
# 对文件进行解码并且对图片作归一化处理
raw_input = decode(contents)
raw_input = tf.image.convert_image_dtype(raw_input, dtype=tf.float32) # 归一化处理
# 判断两个值知否相等，如果不等抛出异常
assertion = tf.assert_equal(tf.shape(raw_input)[2], 3, message="image does not have 3 channels")
'''
对于control_dependencies这个管理器，只有当里面的操作是一个op时，才会生效，也就是先执行传入的
参数op，再执行里面的op。如果里面的操作不是定义的op，图中就不会形成一个节点，这样该管理器就失效了。
tf.identity是返回一个一模一样新的tensor的op，这会增加一个新节点到gragh中，这时control_dependencies就会生效.
'''
with tf.control_dependencies([assertion]):
raw_input = tf.identity(raw_input)
raw_input.set_shape([None, None, 3])
# 图像值由[0,1]--->[-1, 1]
width = tf.shape(raw_input)[1] # [height, width, channels]
a_images = preprocess(raw_input[:,:width//2,:]) # 256*256*3
b_images = preprocess(raw_input[:,width//2:,:]) # 256*256*3
# 这里的which_direction为：BtoA
if which_direction == "AtoB":
inputs, targets = [a_images, b_images]
elif which_direction == "BtoA":
inputs, targets = [b_images, a_images]
else:
raise Exception("invalid direction")
# synchronize seed for image operations so that we do the same operations to both
# input and output images
seed = random.randint(0, 2**31 - 1)
# 图像预处理，翻转、改变形状
with tf.name_scope("input_images"):
input_images = transform(inputs)
with tf.name_scope("target_images"):
target_images = transform(targets)
# 获得输入图像、目标图像的batch块
paths_batch, inputs_batch, targets_batch = tf.train.batch([paths, input_images, target_images], batch_size=batch_size)
steps_per_epoch = int(math.ceil(len(input_paths) / batch_size))
return Examples(
paths=paths_batch, # 输入的文件名块
inputs=inputs_batch, # 输入的图像块
targets=targets_batch, # 目标图像块
count=len(input_paths), # 数据集的大小
steps_per_epoch=steps_per_epoch, # batch的个数
)
# 图像预处理，翻转、改变形状
def transform(image):
r = image
if flip:
r = tf.image.random_flip_left_right(r, seed=seed)
# area produces a nice downscaling, but does nearest neighbor for upscaling
# assume we're going to be doing downscaling here
r = tf.image.resize_images(r, [scale_size, scale_size], method=tf.image.ResizeMethod.AREA)
offset = tf.cast(tf.floor(tf.random_uniform([2], 0, scale_size - CROP_SIZE + 1, seed=seed)), dtype=tf.int32)
if scale_size > CROP_SIZE:
r = tf.image.crop_to_bounding_box(r, offset[0], offset[1], CROP_SIZE, CROP_SIZE)
elif scale_size < CROP_SIZE:
raise Exception("scale size cannot be less than crop size")
return r
#创建生成器，这是一个编码解码器的变种，输入输出均为：256*256*3, 像素值为[-1,1]
def create_generator(generator_inputs, generator_outputs_channels):
layers = []
# encoder_1: [batch, 256, 256, in_channels] => [batch, 128, 128, ngf]
with tf.variable_scope("encoder_1"):
output = gen_conv(generator_inputs, ngf) # ngf为第一个卷积层的卷积核核数量，默认为 64
layers.append(output)
layer_specs = [
ngf * 2, # encoder_2: [batch, 128, 128, ngf] => [batch, 64, 64, ngf * 2]
ngf * 4, # encoder_3: [batch, 64, 64, ngf * 2] => [batch, 32, 32, ngf * 4]
ngf * 8, # encoder_4: [batch, 32, 32, ngf * 4] => [batch, 16, 16, ngf * 8]
ngf * 8, # encoder_5: [batch, 16, 16, ngf * 8] => [batch, 8, 8, ngf * 8]
ngf * 8, # encoder_6: [batch, 8, 8, ngf * 8] => [batch, 4, 4, ngf * 8]
ngf * 8, # encoder_7: [batch, 4, 4, ngf * 8] => [batch, 2, 2, ngf * 8]
ngf * 8, # encoder_8: [batch, 2, 2, ngf * 8] => [batch, 1, 1, ngf * 8]
]
# 卷积的编码器
for out_channels in layer_specs:
with tf.variable_scope("encoder_%d" % (len(layers) + 1)):
# 对最后一层使用激活函数
rectified = lrelu(layers[-1], 0.2)
# [batch, in_height, in_width, in_channels] => [batch, in_height/2, in_width/2, out_channels]
convolved = gen_conv(rectified, out_channels)
output = batchnorm(convolved)
layers.append(output)
layer_specs = [
(ngf * 8, 0.5), # decoder_8: [batch, 1, 1, ngf * 8] => [batch, 2, 2, ngf * 8 * 2]
(ngf * 8, 0.5), # decoder_7: [batch, 2, 2, ngf * 8 * 2] => [batch, 4, 4, ngf * 8 * 2]
(ngf * 8, 0.5), # decoder_6: [batch, 4, 4, ngf * 8 * 2] => [batch, 8, 8, ngf * 8 * 2]
(ngf * 8, 0.0), # decoder_5: [batch, 8, 8, ngf * 8 * 2] => [batch, 16, 16, ngf * 8 * 2]
(ngf * 4, 0.0), # decoder_4: [batch, 16, 16, ngf * 8 * 2] => [batch, 32, 32, ngf * 4 * 2]
(ngf * 2, 0.0), # decoder_3: [batch, 32, 32, ngf * 4 * 2] => [batch, 64, 64, ngf * 2 * 2]
(ngf, 0.0), # decoder_2: [batch, 64, 64, ngf * 2 * 2] => [batch, 128, 128, ngf * 2]
]
# 卷积的解码器
num_encoder_layers = len(layers) # 8
for decoder_layer, (out_channels, dropout) in enumerate(layer_specs):
skip_layer = num_encoder_layers - decoder_layer - 1
with tf.variable_scope("decoder_%d" % (skip_layer + 1)):
if decoder_layer == 0:
# first decoder layer doesn't have skip connections
# since it is directly connected to the skip_layer
input = layers[-1]
else:
input = tf.concat([layers[-1], layers[skip_layer]], axis=3)
rectified = tf.nn.relu(input)
# [batch, in_height, in_width, in_channels] => [batch, in_height*2, in_width*2, out_channels]
output = gen_deconv(rectified, out_channels)
output = batchnorm(output)
if dropout > 0.0:
output = tf.nn.dropout(output, keep_prob=1 - dropout)
layers.append(output)
# decoder_1: [batch, 128, 128, ngf * 2] => [batch, 256, 256, generator_outputs_channels]
with tf.variable_scope("decoder_1"):
input = tf.concat([layers[-1], layers[0]], axis=3)
rectified = tf.nn.relu(input)
output = gen_deconv(rectified, generator_outputs_channels)
output = tf.tanh(output)
layers.append(output)
return layers[-1]
# 创建判别器，输入生成的图像和真实的图像：两个[batch,256,256,3],元素值值[-1,1]，输出:[batch,30,30,1],元素值为概率
def create_discriminator(discrim_inputs, discrim_targets):
n_layers = 3
layers = []
# 2x [batch, height, width, in_channels] => [batch, height, width, in_channels * 2]
input = tf.concat([discrim_inputs, discrim_targets], axis=3)
# layer_1: [batch, 256, 256, in_channels * 2] => [batch, 128, 128, ndf]
with tf.variable_scope("layer_1"):
convolved = discrim_conv(input, ndf, stride=2)
rectified = lrelu(convolved, 0.2)
layers.append(rectified)
# layer_2: [batch, 128, 128, ndf] => [batch, 64, 64, ndf * 2]
# layer_3: [batch, 64, 64, ndf * 2] => [batch, 32, 32, ndf * 4]
# layer_4: [batch, 32, 32, ndf * 4] => [batch, 31, 31, ndf * 8]
for i in range(n_layers):
with tf.variable_scope("layer_%d" % (len(layers) + 1)):
out_channels = ndf * min(2**(i+1), 8)
stride = 1 if i == n_layers - 1 else 2 # last layer here has stride 1
convolved = discrim_conv(layers[-1], out_channels, stride=stride)
normalized = batchnorm(convolved)
rectified = lrelu(normalized, 0.2)
layers.append(rectified)
# layer_5: [batch, 31, 31, ndf * 8] => [batch, 30, 30, 1]
with tf.variable_scope("layer_%d" % (len(layers) + 1)):
convolved = discrim_conv(rectified, out_channels=1, stride=1)
output = tf.sigmoid(convolved)
layers.append(output)
return layers[-1]
# 创建Pix2Pix模型，inputs和targets形状为：[batch_size, height, width, channels]
def create_model(inputs, targets):
with tf.variable_scope("generator"):
out_channels = int(targets.get_shape()[-1])
outputs = create_generator(inputs, out_channels)
# create two copies of discriminator, one for real pairs and one for fake pairs
# they share the same underlying variables
with tf.name_scope("real_discriminator"):
with tf.variable_scope("discriminator"):
# 2x [batch, height, width, channels] => [batch, 30, 30, 1]
predict_real = create_discriminator(inputs, targets) # 条件变量图像和真实图像
with tf.name_scope("fake_discriminator"):
with tf.variable_scope("discriminator", reuse=True):
# 2x [batch, height, width, channels] => [batch, 30, 30, 1]
predict_fake = create_discriminator(inputs, outputs) # 条件变量图像和生成的图像
# 判别器的损失，判别器希望V(G,D)尽可能大
with tf.name_scope("discriminator_loss"):
# minimizing -tf.log will try to get inputs to 1
# predict_real => 1
# predict_fake => 0
discrim_loss = tf.reduce_mean(-(tf.log(predict_real + EPS) + tf.log(1 - predict_fake + EPS)))
# 生成器的损失，生成器希望V(G,D)尽可能小
with tf.name_scope("generator_loss"):
# predict_fake => 1
# abs(targets - outputs) => 0
gen_loss_GAN = tf.reduce_mean(-tf.log(predict_fake + EPS))
gen_loss_L1 = tf.reduce_mean(tf.abs(targets - outputs))
gen_loss = gen_loss_GAN * gan_weight + gen_loss_L1 * l1_weight
# 判别器训练
with tf.name_scope("discriminator_train"):
# 判别器需要优化的参数
discrim_tvars = [var for var in tf.trainable_variables() if var.name.startswith("discriminator")]
# 优化器定义
discrim_optim = tf.train.AdamOptimizer(lr, beta1)
# 计算损失函数对优化参数的梯度
discrim_grads_and_vars = discrim_optim.compute_gradients(discrim_loss, var_list=discrim_tvars)
# 更新该梯度所对应的参数的状态，返回一个op
discrim_train = discrim_optim.apply_gradients(discrim_grads_and_vars)
# 生成器训练
with tf.name_scope("generator_train"):
with tf.control_dependencies([discrim_train]):
# 生成器需要优化的参数列表
gen_tvars = [var for var in tf.trainable_variables() if var.name.startswith("generator")]
# 定义优化器
gen_optim = tf.train.AdamOptimizer(lr, beta1)
# 计算需要优化的参数的梯度
gen_grads_and_vars = gen_optim.compute_gradients(gen_loss, var_list=gen_tvars)
# 更新该梯度所对应的参数的状态，返回一个op
gen_train = gen_optim.apply_gradients(gen_grads_and_vars)
'''
在采用随机梯度下降算法训练神经网络时，使用 tf.train.ExponentialMovingAverage 滑动平均操作的意义在于
提高模型在测试数据上的健壮性（robustness）。tensorflow 下的 tf.train.ExponentialMovingAverage 需要
提供一个衰减率（decay）。该衰减率用于控制模型更新的速度。该衰减率用于控制模型更新的速度，
ExponentialMovingAverage 对每一个（待更新训练学习的）变量（variable）都会维护一个影子变量
（shadow variable）。影子变量的初始值就是这个变量的初始值，
shadow_variable=decay×shadow_variable+(1−decay)×variable
'''
ema = tf.train.ExponentialMovingAverage(decay=0.99)
update_losses = ema.apply([discrim_loss, gen_loss_GAN, gen_loss_L1])
#
global_step = tf.train.get_or_create_global_step()
incr_global_step = tf.assign(global_step, global_step+1)
return Model(
predict_real=predict_real, # 条件变量(输入图像)和真实图像之间的概率值，形状为；[batch,30,30,1]
predict_fake=predict_fake, # 条件变量(输入图像)和生成图像之间的概率值，形状为；[batch,30,30,1]
discrim_loss=ema.average(discrim_loss), # 判别器损失
discrim_grads_and_vars=discrim_grads_and_vars, # 判别器需要优化的参数和对应的梯度
gen_loss_GAN=ema.average(gen_loss_GAN), # 生成器的损失
gen_loss_L1=ema.average(gen_loss_L1), # 生成器的 L1损失
gen_grads_and_vars=gen_grads_and_vars, # 生成器需要优化的参数和对应的梯度
outputs=outputs, # 生成器生成的图片
train=tf.group(update_losses, incr_global_step, gen_train), # 打包需要run的操作op
)
# 保存图像
def save_images(output_dir,fetches, step=None):
image_dir = os.path.join(output_dir, "images")
if not os.path.exists(image_dir):
os.makedirs(image_dir)
filesets = []
for i, in_path in enumerate(fetches["paths"]):
name, _ = os.path.splitext(os.path.basename(in_path.decode("utf8")))
fileset = {"name": name, "step": step}
for kind in ["inputs", "outputs", "targets"]:
filename = name + "-" + kind + ".png"
if step is not None:
filename = "%08d-%s" % (step, filename)
fileset[kind] = filename
out_path = os.path.join(image_dir, filename)
contents = fetches[kind][i]
with open(out_path, "wb") as f:
f.write(contents)
filesets.append(fileset)
return filesets
# 将结果写入HTML网页
def append_index(output_dir,filesets, step=False):
index_path = os.path.join(output_dir, "index.html")
if os.path.exists(index_path):
index = open(index_path, "a")
else:
index = open(index_path, "w")
index.write("<html><body><table><tr>")
if step:
index.write("<th>step</th>")
index.write("<th>name</th><th>input</th><th>output</th><th>target</th></tr>")
for fileset in filesets:
index.write("<tr>")
if step:
index.write("<td>%d</td>" % fileset["step"])
index.write("<td>%s</td>" % fileset["name"])
for kind in ["inputs", "outputs", "targets"]:
index.write("<td><img src='images/%s'></td>" % fileset[kind])
index.write("</tr>")
return index_path
# 转变图像的尺寸、并且将[0,1]--->[0,255]
def convert(image):
if aspect_ratio != 1.0:
# upscale to correct aspect ratio
size = [CROP_SIZE, int(round(CROP_SIZE * a.aspect_ratio))]
image = tf.image.resize_images(image, size=size, method=tf.image.ResizeMethod.BICUBIC)
# 将数据的类型转换为8位无符号整型
return tf.image.convert_image_dtype(image, dtype=tf.uint8, saturate=True)
# 主函数
def train():
# 设置随机数种子的值
global seed
if seed is None:
seed = random.randint(0, 2**31 - 1)
tf.set_random_seed(seed)
np.random.seed(seed)
random.seed(seed)
# 创建目录
if not os.path.exists(train_output_dir):
os.makedirs(train_output_dir)
# 加载数据集，得到输入数据和目标数据并把范围变为 :[-1,1]
examples = load_examples(train_input_dir)
print("load successful ! examples count = %d" % examples.count)
# 创建模型，inputs和targets是：[batch_size, height, width, channels]
# 返回值：
model = create_model(examples.inputs, examples.targets)
print ("create model successful!")
#图像处理[-1, 1] => [0, 1]
inputs = deprocess(examples.inputs)
targets = deprocess(examples.targets)
outputs = deprocess(model.outputs)
# 把[0,1]的像素点转为RGB值：[0,255]
with tf.name_scope("convert_inputs"):
converted_inputs = convert(inputs)
with tf.name_scope("convert_targets"):
converted_targets = convert(targets)
with tf.name_scope("convert_outputs"):
converted_outputs = convert(outputs)
# 对图像进行编码以便于保存
with tf.name_scope("encode_images"):
display_fetches = {
"paths": examples.paths,
# tf.map_fn接受一个函数对象和集合，用函数对集合中每个元素分别处理
"inputs": tf.map_fn(tf.image.encode_png, converted_inputs, dtype=tf.string, name="input_pngs"),
"targets": tf.map_fn(tf.image.encode_png, converted_targets, dtype=tf.string, name="target_pngs"),
"outputs": tf.map_fn(tf.image.encode_png, converted_outputs, dtype=tf.string, name="output_pngs"),
}
with tf.name_scope("parameter_count"):
parameter_count = tf.reduce_sum([tf.reduce_prod(tf.shape(v)) for v in tf.trainable_variables()])
# 只保存最新一个checkpoint
saver = tf.train.Saver(max_to_keep=20)
init=tf.global_variables_initializer()
with tf.Session() as sess:
sess.run(init)
print("parameter_count =", sess.run(parameter_count))
if max_epochs is not None:
max_steps = examples.steps_per_epoch * max_epochs # 400X200=8000
# 因为是从文件中读取数据，所以需要启动start_queue_runners()
# 这个函数将会启动输入管道的线程，填充样本到队列中，以便出队操作可以从队列中拿到样本。
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(coord=coord)
# 运行训练集
print ("begin trainning......")
print ("max_steps:",max_steps)
start = time.time()
for step in range(max_steps):
def should(freq):
return freq > 0 and ((step + 1) % freq == 0 or step == max_steps - 1)
print ("step:",step)
# 定义一个需要run的所有操作的字典
fetches = {
"train": model.train
}
# progress_freq为 50，每50次计算一次三个损失，显示进度
if should(progress_freq):
fetches["discrim_loss"] = model.discrim_loss
fetches["gen_loss_GAN"] = model.gen_loss_GAN
fetches["gen_loss_L1"] = model.gen_loss_L1
# display_freq为 50，每50次保存一次输入、目标、输出的图像
if should(display_freq):
fetches["display"] = display_fetches
# 运行各种操作，
results = sess.run(fetches)
# display_freq为 50，每50次保存输入、目标、输出的图像
if should(display_freq):
print("saving display images")
filesets = save_images(train_output_dir,results["display"], step=step)
append_index(train_output_dir,filesets, step=True)
# progress_freq为 50，每50次打印一次三种损失的大小，显示进度
if should(progress_freq):
# global_step will have the correct step count if we resume from a checkpoint
train_epoch = math.ceil(step/examples.steps_per_epoch)
train_step = (step - 1) % examples.steps_per_epoch + 1
rate = (step + 1) * batch_size / (time.time() - start)
remaining = (max_steps - step) * batch_size / rate
print("progress epoch %d step %d image/sec %0.1f remaining %dm" % (train_epoch, train_step, rate, remaining / 60))
print("discrim_loss", results["discrim_loss"])
print("gen_loss_GAN", results["gen_loss_GAN"])
print("gen_loss_L1", results["gen_loss_L1"])
# save_freq为500，每500次保存一次模型
if should(save_freq):
print("saving model")
saver.save(sess, os.path.join(train_output_dir, "model"), global_step=step)
# 测试
def test():
# 设置随机数种子的值
global seed
if seed is None:
seed = random.randint(0, 2**31 - 1)
tf.set_random_seed(seed)
np.random.seed(seed)
random.seed(seed)
# 创建目录
if not os.path.exists(test_output_dir):
os.makedirs(test_output_dir)
if checkpoint is None:
raise Exception("checkpoint required for test mode")
# disable these features in test mode
scale_size = CROP_SIZE
flip = False
# 加载数据集，得到输入数据和目标数据
examples = load_examples(test_input_dir)
print("load successful ! examples count = %d" % examples.count)
# 创建模型，inputs和targets是：[batch_size, height, width, channels]
model = create_model(examples.inputs, examples.targets)
print ("create model successful!")
#图像处理[-1, 1] => [0, 1]
inputs = deprocess(examples.inputs)
targets = deprocess(examples.targets)
outputs = deprocess(model.outputs)
# 把[0,1]的像素点转为RGB值：[0,255]
with tf.name_scope("convert_inputs"):
converted_inputs = convert(inputs)
with tf.name_scope("convert_targets"):
converted_targets = convert(targets)
with tf.name_scope("convert_outputs"):
converted_outputs = convert(outputs)
# 对图像进行编码以便于保存
with tf.name_scope("encode_images"):
display_fetches = {
"paths": examples.paths,
# tf.map_fn接受一个函数对象和集合，用函数对集合中每个元素分别处理
"inputs": tf.map_fn(tf.image.encode_png, converted_inputs, dtype=tf.string, name="input_pngs"),
"targets": tf.map_fn(tf.image.encode_png, converted_targets, dtype=tf.string, name="target_pngs"),
"outputs": tf.map_fn(tf.image.encode_png, converted_outputs, dtype=tf.string, name="output_pngs"),
}
sess=tf.InteractiveSession()
saver = tf.train.Saver(max_to_keep=1)
ckpt=tf.train.get_checkpoint_state(checkpoint)
saver.restore(sess, ckpt.model_checkpoint_path)
start = time.time()
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(coord=coord)
for step in range(examples.count):
results = sess.run(display_fetches)
filesets = save_images(test_output_dir,results)
for i, f in enumerate(filesets):
print("evaluated image", f["name"])
index_path = append_index(test_output_dir,filesets)
print("wrote index at", index_path)
print("rate", (time.time() - start) / max_steps)
if __name__ == '__main__':
# test()
train()