Skip to content
体验新版
项目
组织
正在加载...
登录
切换导航
打开侧边栏
PaddlePaddle
models
提交
23d450e9
M
models
项目概览
PaddlePaddle
/
models
大约 1 年 前同步成功
通知
222
Star
6828
Fork
2962
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
602
列表
看板
标记
里程碑
合并请求
255
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
M
models
项目概览
项目概览
详情
发布
仓库
仓库
文件
提交
分支
标签
贡献者
分支图
比较
Issue
602
Issue
602
列表
看板
标记
里程碑
合并请求
255
合并请求
255
Pages
分析
分析
仓库分析
DevOps
Wiki
0
Wiki
成员
成员
收起侧边栏
关闭侧边栏
动态
分支图
创建新Issue
提交
Issue看板
体验新版 GitCode,发现更多精彩内容 >>
未验证
提交
23d450e9
编写于
7月 17, 2019
作者:
L
lvmengsi
提交者:
GitHub
7月 17, 2019
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Cherry pick gan0717 (#2831)
update cherry pick
上级
c399ca44
变更
15
隐藏空白更改
内联
并排
Showing
15 changed file
with
116 addition
and
67 deletion
+116
-67
PaddleCV/PaddleGAN/README.md
PaddleCV/PaddleGAN/README.md
+26
-13
PaddleCV/PaddleGAN/images/attgan.jpg
PaddleCV/PaddleGAN/images/attgan.jpg
+0
-0
PaddleCV/PaddleGAN/images/stargan.jpg
PaddleCV/PaddleGAN/images/stargan.jpg
+0
-0
PaddleCV/PaddleGAN/images/stgan.jpg
PaddleCV/PaddleGAN/images/stgan.jpg
+0
-0
PaddleCV/PaddleGAN/infer.py
PaddleCV/PaddleGAN/infer.py
+14
-11
PaddleCV/PaddleGAN/network/AttGAN_network.py
PaddleCV/PaddleGAN/network/AttGAN_network.py
+3
-2
PaddleCV/PaddleGAN/network/CGAN_network.py
PaddleCV/PaddleGAN/network/CGAN_network.py
+7
-3
PaddleCV/PaddleGAN/network/DCGAN_network.py
PaddleCV/PaddleGAN/network/DCGAN_network.py
+7
-3
PaddleCV/PaddleGAN/network/STGAN_network.py
PaddleCV/PaddleGAN/network/STGAN_network.py
+9
-8
PaddleCV/PaddleGAN/network/base_network.py
PaddleCV/PaddleGAN/network/base_network.py
+1
-1
PaddleCV/PaddleGAN/scripts/run_attgan.sh
PaddleCV/PaddleGAN/scripts/run_attgan.sh
+1
-1
PaddleCV/PaddleGAN/scripts/run_stgan.sh
PaddleCV/PaddleGAN/scripts/run_stgan.sh
+1
-1
PaddleCV/PaddleGAN/trainer/AttGAN.py
PaddleCV/PaddleGAN/trainer/AttGAN.py
+6
-0
PaddleCV/PaddleGAN/trainer/STGAN.py
PaddleCV/PaddleGAN/trainer/STGAN.py
+7
-1
PaddleCV/PaddleGAN/util/utility.py
PaddleCV/PaddleGAN/util/utility.py
+34
-23
未找到文件。
PaddleCV/PaddleGAN/README.md
浏览文件 @
23d450e9
...
@@ -18,9 +18,11 @@
...
@@ -18,9 +18,11 @@
本图像生成模型库包含CGAN
\[
[
3
](
#参考文献
)
\]
, DCGAN
\[
[
4
](
#参考文献
)
\]
, Pix2Pix
\[
[
5
](
#参考文献
)
\]
, CycleGAN
\[
[
6
](
#参考文献
)
\]
, StarGAN
\[
[
7
](
#参考文献
)
\]
, AttGAN
\[
[
8
](
#参考文献
)
\]
, STGAN
\[
[
9
](
#参考文献
)
\]
。
本图像生成模型库包含CGAN
\[
[
3
](
#参考文献
)
\]
, DCGAN
\[
[
4
](
#参考文献
)
\]
, Pix2Pix
\[
[
5
](
#参考文献
)
\]
, CycleGAN
\[
[
6
](
#参考文献
)
\]
, StarGAN
\[
[
7
](
#参考文献
)
\]
, AttGAN
\[
[
8
](
#参考文献
)
\]
, STGAN
\[
[
9
](
#参考文献
)
\]
。
注意:
注意:
1.
AttGAN和STGAN的网络结构中,判别器去掉了instance norm。
1.
StarGAN,AttGAN和STGAN由于梯度惩罚所需的操作目前只支持GPU,需使用GPU训练。
2.
StarGAN,AttGAN和STGAN由于梯度惩罚所需的操作目前只支持GPU,需使用GPU训练。
2.
CGAN和DCGAN仅支持多batch size训练。
3.
CGAN和DCGAN两个模型训练使用的数据集为MNIST数据集;StarGAN,AttGAN和STGAN的数据集为CelebA数据集,测试集列表(test_list)和下载到的list文件格式相同,即包含测试集数量,属性列表,想要进行测试的图片和标签。Pix2Pix和CycleGAN支持的数据集可以参考download.py中的cycle_pix_dataset。
4.
PaddlePaddle1.5.1及之前的版本不支持在AttGAN和STGAN模型里的判别器加上的instance norm。如果要在判别器中加上instance norm,请源码编译develop分支并安装。
图像生成模型库库的目录结构如下:
图像生成模型库库的目录结构如下:
```
```
...
@@ -58,7 +60,7 @@
...
@@ -58,7 +60,7 @@
### 安装说明
### 安装说明
**安装[PaddlePaddle](https://github.com/PaddlePaddle/Paddle):**
**安装[PaddlePaddle](https://github.com/PaddlePaddle/Paddle):**
在当前目录下运行样例代码需要PadddlePaddle Fluid的v.1.5或以上的版本。如果你的运行环境中的PaddlePaddle低于此版本,请根据
[
安装文档
](
http
://paddlepaddle.org/documentation/docs/zh/1.4
/beginners_guide/install/index_cn.html
)
中的说明来更新PaddlePaddle。
在当前目录下运行样例代码需要PadddlePaddle Fluid的v.1.5或以上的版本。如果你的运行环境中的PaddlePaddle低于此版本,请根据
[
安装文档
](
http
s://www.paddlepaddle.org.cn/documentation/docs/zh/1.5
/beginners_guide/install/index_cn.html
)
中的说明来更新PaddlePaddle。
### 任务简介
### 任务简介
...
@@ -141,11 +143,19 @@ Pix2Pix和CycleGAN的效果图
...
@@ -141,11 +143,19 @@ Pix2Pix和CycleGAN的效果图
StarGAN,AttGAN和STGAN的效果如图所示:
StarGAN,AttGAN和STGAN的效果如图所示:
<p
align=
"center"
>
<p
align=
"center"
>
<img
src=
"images/
female_stargan_attgan_stgan.png"
width=
"65
0"
/><br
/>
<img
src=
"images/
stargan.jpg"
width=
"50
0"
/><br
/>
StarGAN
,AttGAN和STGAN的效果图
StarGAN
的效果图(图片属性分别为:origial image, Black hair, Blond Hair, Brown Hair, Male, Young)
</p>
</p>
<p
align=
"center"
>
<img
src=
"images/attgan.jpg"
width=
"1250"
/><br
/>
AttGAN的效果图(图片属性分别为:original image, Bald, Bangs, Black Hair, Blond Hair, Brown Hair, Bushy Eyebrows, Eyeglasses, Male, Mouth Slightly Open, Mustache, No Beard, Pale Skin, Young)
</p>
<p
align=
"center"
>
<img
src=
"images/stgan.jpg"
width=
"1250"
/><br
/>
STGAN的效果图(图片属性分别为:original image, Bald, Bangs, Black Hair, Blond Hair, Brown Hair, Bushy Eyebrows, Eyeglasses, Male, Mouth Slightly Open, Mustache, No Beard, Pale Skin, Young)
</p>
-
每个GAN都给出了一份测试示例,放在scripts文件夹内,用户可以直接运行测试脚本得到测试结果。
-
每个GAN都给出了一份测试示例,放在scripts文件夹内,用户可以直接运行测试脚本得到测试结果。
...
@@ -181,7 +191,7 @@ STGAN只输入有变化的标签,引入GRU结构,更好的选择变化的属
...
@@ -181,7 +191,7 @@ STGAN只输入有变化的标签,引入GRU结构,更好的选择变化的属
### 模型概览
### 模型概览
-
Pix2Pix由一个生成网络和一个判别网络组成。生成网络中编码部分的网络结构都是采用
`convolution-batch norm-ReLU`
作为基础结构,解码部分的网络结构由
`transpose convolution-batch norm-ReLU`
组成,判别网络基本是由
`convolution-norm-leaky_ReLU`
作为基础结构,详细的网络结构可以查看
`network/Pix2pix_network.py`
文件。生成网络提供两种可选的网络结构:Unet网络结构和普通的encoder-decoder网络结构。网络利用损失函数学习从输入图像到输出图像的映射,生成网络损失函数由
CGAN的损失函数和L1损失函数组成,判别网络损失函数由C
GAN的损失函数组成。生成器的网络结构如下图所示:
-
Pix2Pix由一个生成网络和一个判别网络组成。生成网络中编码部分的网络结构都是采用
`convolution-batch norm-ReLU`
作为基础结构,解码部分的网络结构由
`transpose convolution-batch norm-ReLU`
组成,判别网络基本是由
`convolution-norm-leaky_ReLU`
作为基础结构,详细的网络结构可以查看
`network/Pix2pix_network.py`
文件。生成网络提供两种可选的网络结构:Unet网络结构和普通的encoder-decoder网络结构。网络利用损失函数学习从输入图像到输出图像的映射,生成网络损失函数由
GAN的损失函数和L1损失函数组成,判别网络损失函数由
GAN的损失函数组成。生成器的网络结构如下图所示:
<p
align=
"center"
>
<p
align=
"center"
>
<img
src=
"images/pix2pix_gen.png"
width=
"550"
/><br
/>
<img
src=
"images/pix2pix_gen.png"
width=
"550"
/><br
/>
...
@@ -189,7 +199,7 @@ Pix2Pix生成网络结构图[5]
...
@@ -189,7 +199,7 @@ Pix2Pix生成网络结构图[5]
</p>
</p>
-
CycleGAN由两个生成网络和两个判别网络组成,生成网络A是输入A类风格的图片输出B类风格的图片,生成网络B是输入B类风格的图片输出A类风格的图片。生成网络中编码部分的网络结构都是采用
`convolution-norm-ReLU`
作为基础结构,解码部分的网络结构由
`transpose convolution-norm-ReLU`
组成,判别网络基本是由
`convolution-norm-leaky_ReLU`
作为基础结构,详细的网络结构可以查看
`network/CycleGAN_network.py`
文件。生成网络提供两种可选的网络结构:Unet网络结构和普通的encoder-decoder网络结构。生成网络损失函数由
CGAN的损失函数,重构损失和自身损失组成,判别网络的损失函数由C
GAN的损失函数组成。
-
CycleGAN由两个生成网络和两个判别网络组成,生成网络A是输入A类风格的图片输出B类风格的图片,生成网络B是输入B类风格的图片输出A类风格的图片。生成网络中编码部分的网络结构都是采用
`convolution-norm-ReLU`
作为基础结构,解码部分的网络结构由
`transpose convolution-norm-ReLU`
组成,判别网络基本是由
`convolution-norm-leaky_ReLU`
作为基础结构,详细的网络结构可以查看
`network/CycleGAN_network.py`
文件。生成网络提供两种可选的网络结构:Unet网络结构和普通的encoder-decoder网络结构。生成网络损失函数由
LSGAN的损失函数,重构损失和自身损失组成,判别网络的损失函数由LS
GAN的损失函数组成。
<p
align=
"center"
>
<p
align=
"center"
>
<img
src=
"images/pix2pix_gen.png"
width=
"550"
/><br
/>
<img
src=
"images/pix2pix_gen.png"
width=
"550"
/><br
/>
...
@@ -197,7 +207,7 @@ CycleGAN生成网络结构图[5]
...
@@ -197,7 +207,7 @@ CycleGAN生成网络结构图[5]
</p>
</p>
-
StarGAN中生成网络的编码部分主要由
`convolution-instance norm-ReLU`
组成,解码部分主要由
`transpose convolution-norm-ReLU`
组成,判别网络主要由
`convolution-leaky_ReLU`
组成,详细网络结构可以查看
`network/StarGAN_network.py`
文件。生成网络的损失函数是由
C
GAN的损失函数,重构损失和分类损失组成,判别网络的损失函数由预测损失,分类损失和梯度惩罚损失组成。
-
StarGAN中生成网络的编码部分主要由
`convolution-instance norm-ReLU`
组成,解码部分主要由
`transpose convolution-norm-ReLU`
组成,判别网络主要由
`convolution-leaky_ReLU`
组成,详细网络结构可以查看
`network/StarGAN_network.py`
文件。生成网络的损失函数是由
W
GAN的损失函数,重构损失和分类损失组成,判别网络的损失函数由预测损失,分类损失和梯度惩罚损失组成。
<p
align=
"center"
>
<p
align=
"center"
>
<img
src=
"images/stargan_gen.png"
width=
350
/>
<img
src=
"images/stargan_gen.png"
width=
350
/>
...
@@ -207,7 +217,7 @@ StarGAN的生成网络结构[左]和判别网络结构[右] [7]
...
@@ -207,7 +217,7 @@ StarGAN的生成网络结构[左]和判别网络结构[右] [7]
-
AttGAN中生成网络的编码部分主要由
`convolution-instance norm-ReLU`
组成,解码部分由
`transpose convolution-norm-ReLU`
组成,判别网络主要由
`convolution-leaky_ReLU`
组成,详细网络结构可以查看
`network/AttGAN_network.py`
文件。生成网络的损失函数是由
C
GAN的损失函数,重构损失和分类损失组成,判别网络的损失函数由预测损失,分类损失和梯度惩罚损失组成。
-
AttGAN中生成网络的编码部分主要由
`convolution-instance norm-ReLU`
组成,解码部分由
`transpose convolution-norm-ReLU`
组成,判别网络主要由
`convolution-leaky_ReLU`
组成,详细网络结构可以查看
`network/AttGAN_network.py`
文件。生成网络的损失函数是由
W
GAN的损失函数,重构损失和分类损失组成,判别网络的损失函数由预测损失,分类损失和梯度惩罚损失组成。
<p
align=
"center"
>
<p
align=
"center"
>
<img
src=
"images/attgan_net.png"
width=
800
/>
<br
/>
<img
src=
"images/attgan_net.png"
width=
800
/>
<br
/>
...
@@ -215,7 +225,7 @@ AttGAN的网络结构[8]
...
@@ -215,7 +225,7 @@ AttGAN的网络结构[8]
</p>
</p>
-
STGAN中生成网络再编码器和解码器之间加入Selective Transfer Units
\(
STU
\)
,有选择的转换编码网络,从而更好的适配解码网络。生成网络中的编码网络主要由
`convolution-instance norm-ReLU`
组成,解码网络主要由
`transpose convolution-norm-leaky_ReLU`
组成,判别网络主要由
`convolution-leaky_ReLU`
组成,详细网络结构可以查看
`network/STGAN_network.py`
文件。生成网络的损失函数是由
C
GAN的损失函数,重构损失和分类损失组成,判别网络的损失函数由预测损失,分类损失和梯度惩罚损失组成。
-
STGAN中生成网络再编码器和解码器之间加入Selective Transfer Units
\(
STU
\)
,有选择的转换编码网络,从而更好的适配解码网络。生成网络中的编码网络主要由
`convolution-instance norm-ReLU`
组成,解码网络主要由
`transpose convolution-norm-leaky_ReLU`
组成,判别网络主要由
`convolution-leaky_ReLU`
组成,详细网络结构可以查看
`network/STGAN_network.py`
文件。生成网络的损失函数是由
W
GAN的损失函数,重构损失和分类损失组成,判别网络的损失函数由预测损失,分类损失和梯度惩罚损失组成。
<p
align=
"center"
>
<p
align=
"center"
>
<img
src=
"images/stgan_net.png"
width=
800
/>
<br
/>
<img
src=
"images/stgan_net.png"
width=
800
/>
<br
/>
...
@@ -228,17 +238,20 @@ STGAN的网络结构[9]
...
@@ -228,17 +238,20 @@ STGAN的网络结构[9]
## FAQ
## FAQ
**Q:**
StarGAN/AttGAN/STGAN中属性没有变化,为什么?
**Q:**
StarGAN/AttGAN/STGAN中属性没有变化,为什么?
**A:**
查看是否所有的标签都转换对了。
**A:**
查看是否所有的标签都转换对了。
**Q:**
预测结果不正常,是怎么回事?
**Q:**
预测结果不正常,是怎么回事?
**A:**
某些GAN预测的时候batch_norm的设置需要和训练的时候行为一致,查看模型库中相应的GAN中预测时batch_norm的行为和自己模型中的预测时batch_norm的
**A:**
某些GAN预测的时候batch_norm的设置需要和训练的时候行为一致,查看模型库中相应的GAN中预测时batch_norm的行为和自己模型中的预测时batch_norm的
行为是否一致。
行为是否一致。
**Q:**
为什么STGAN和ATTGAN中变男性得到的预测结果是变女性呢?
**Q:**
为什么STGAN和ATTGAN中变男性得到的预测结果是变女性呢?
**A:**
这是由于预测时标签的设置,目标标签是基于原本的标签进行改变,比如原本图片是男生,预测代码对标签进行转变的时候会自动变成相对立的标签,即女
**A:**
这是由于预测时标签的设置,目标标签是基于原本的标签进行改变,比如原本图片是男生,预测代码对标签进行转变的时候会自动变成相对立的标签,即女
性,所以得到的结果是女生。如果想要原本是男生,转变之后还是男生,保持要转变的标签不变即可。
性,所以得到的结果是女生。如果想要原本是男生,转变之后还是男生,保持要转变的标签不变即可。
**Q:**
如何使用自己的数据集进行训练?
**A:**
对于Pix2Pix来说,只要准备好类似于Cityscapes数据集的不同风格的成对的数据即可。对于CycleGAN来说,只要准备类似于Cityscapes数据集的不同风格的数据即可。对于StarGAN,AttGAN和STGAN来说,除了需要准备类似于CelebA数据集中的图片和标签文件外,还需要把模型中的selected_attrs参数设置为想要改变的目标属性,c_dim参数这是为目标属性的个数。
## 参考论文
## 参考论文
[
1] [Goodfellow, Ian J.; Pouget-Abadie, Jean; Mirza, Mehdi; Xu, Bing; Warde-Farley, David; Ozair, Sherjil; Courville, Aaron; Bengio, Yoshua. Generative Adversarial Networks. 2014. arXiv:1406.2661 [stat.ML].
](
https://arxiv.org/abs/1406.2661
)
[
1] [Goodfellow, Ian J.; Pouget-Abadie, Jean; Mirza, Mehdi; Xu, Bing; Warde-Farley, David; Ozair, Sherjil; Courville, Aaron; Bengio, Yoshua. Generative Adversarial Networks. 2014. arXiv:1406.2661 [stat.ML].
](
https://arxiv.org/abs/1406.2661
)
...
...
PaddleCV/PaddleGAN/images/attgan.jpg
0 → 100644
浏览文件 @
23d450e9
34.6 KB
PaddleCV/PaddleGAN/images/stargan.jpg
0 → 100644
浏览文件 @
23d450e9
16.2 KB
PaddleCV/PaddleGAN/images/stgan.jpg
0 → 100644
浏览文件 @
23d450e9
35.4 KB
PaddleCV/PaddleGAN/infer.py
浏览文件 @
23d450e9
...
@@ -27,7 +27,7 @@ import imageio
...
@@ -27,7 +27,7 @@ import imageio
import
glob
import
glob
from
util.config
import
add_arguments
,
print_arguments
from
util.config
import
add_arguments
,
print_arguments
from
data_reader
import
celeba_reader_creator
from
data_reader
import
celeba_reader_creator
from
util.utility
import
check_attribute_conflict
,
check_gpu
from
util.utility
import
check_attribute_conflict
,
check_gpu
,
save_batch_image
import
copy
import
copy
parser
=
argparse
.
ArgumentParser
(
description
=
__doc__
)
parser
=
argparse
.
ArgumentParser
(
description
=
__doc__
)
...
@@ -118,7 +118,6 @@ def infer(args):
...
@@ -118,7 +118,6 @@ def infer(args):
print
(
args
.
init_model
+
'/'
+
model_name
)
print
(
args
.
init_model
+
'/'
+
model_name
)
fluid
.
io
.
load_persistables
(
exe
,
args
.
init_model
+
"/"
+
model_name
)
fluid
.
io
.
load_persistables
(
exe
,
args
.
init_model
+
"/"
+
model_name
)
print
(
'load params done'
)
print
(
'load params done'
)
if
not
os
.
path
.
exists
(
args
.
output
):
if
not
os
.
path
.
exists
(
args
.
output
):
os
.
makedirs
(
args
.
output
)
os
.
makedirs
(
args
.
output
)
...
@@ -144,7 +143,7 @@ def infer(args):
...
@@ -144,7 +143,7 @@ def infer(args):
tensor_label_trg_
=
fluid
.
LoDTensor
()
tensor_label_trg_
=
fluid
.
LoDTensor
()
tensor_img
.
set
(
real_img
,
place
)
tensor_img
.
set
(
real_img
,
place
)
tensor_label_org
.
set
(
label_org
,
place
)
tensor_label_org
.
set
(
label_org
,
place
)
real_img_temp
=
np
.
squeeze
(
real_img
).
transpose
([
0
,
2
,
3
,
1
]
)
real_img_temp
=
save_batch_image
(
real_img
)
images
=
[
real_img_temp
]
images
=
[
real_img_temp
]
for
i
in
range
(
args
.
c_dim
):
for
i
in
range
(
args
.
c_dim
):
label_trg_tmp
=
copy
.
deepcopy
(
label_trg
)
label_trg_tmp
=
copy
.
deepcopy
(
label_trg
)
...
@@ -152,11 +151,13 @@ def infer(args):
...
@@ -152,11 +151,13 @@ def infer(args):
label_trg_tmp
[
j
][
i
]
=
1.0
-
label_trg_tmp
[
j
][
i
]
label_trg_tmp
[
j
][
i
]
=
1.0
-
label_trg_tmp
[
j
][
i
]
label_trg_tmp
=
check_attribute_conflict
(
label_trg_tmp
=
check_attribute_conflict
(
label_trg_tmp
,
attr_names
[
i
],
attr_names
)
label_trg_tmp
,
attr_names
[
i
],
attr_names
)
label_org_
=
list
(
map
(
lambda
x
:
((
x
*
2
)
-
1
)
*
0.5
,
label_org
))
label_trg_
=
list
(
label_trg_
=
list
(
map
(
lambda
x
:
((
x
*
2
)
-
1
)
*
0.5
,
label_trg_tmp
))
map
(
lambda
x
:
((
x
*
2
)
-
1
)
*
0.5
,
label_trg_tmp
))
for
j
in
range
(
len
(
label_org
)):
if
args
.
model_net
==
'AttGAN'
:
label_trg_
[
j
][
i
]
=
label_trg_
[
j
][
i
]
*
2.0
for
k
in
range
(
len
(
label_org
)):
tensor_label_org_
.
set
(
label_org
,
place
)
label_trg_
[
k
][
i
]
=
label_trg_
[
k
][
i
]
*
2.0
tensor_label_org_
.
set
(
label_org_
,
place
)
tensor_label_trg
.
set
(
label_trg
,
place
)
tensor_label_trg
.
set
(
label_trg
,
place
)
tensor_label_trg_
.
set
(
label_trg_
,
place
)
tensor_label_trg_
.
set
(
label_trg_
,
place
)
out
=
exe
.
run
(
feed
=
{
out
=
exe
.
run
(
feed
=
{
...
@@ -165,10 +166,11 @@ def infer(args):
...
@@ -165,10 +166,11 @@ def infer(args):
"label_trg_"
:
tensor_label_trg_
"label_trg_"
:
tensor_label_trg_
},
},
fetch_list
=
[
fake
.
name
])
fetch_list
=
[
fake
.
name
])
fake_temp
=
np
.
squeeze
(
out
[
0
]).
transpose
([
0
,
2
,
3
,
1
])
fake_temp
=
save_batch_image
(
out
[
0
])
images
.
append
(
fake_temp
)
images
.
append
(
fake_temp
)
images_concat
=
np
.
concatenate
(
images
,
1
)
images_concat
=
np
.
concatenate
(
images
,
1
)
images_concat
=
np
.
concatenate
(
images_concat
,
1
)
if
len
(
label_org
)
>
1
:
images_concat
=
np
.
concatenate
(
images_concat
,
1
)
imageio
.
imwrite
(
args
.
output
+
"/fake_img_"
+
name
[
0
],
(
imageio
.
imwrite
(
args
.
output
+
"/fake_img_"
+
name
[
0
],
(
(
images_concat
+
1
)
*
127.5
).
astype
(
np
.
uint8
))
(
images_concat
+
1
)
*
127.5
).
astype
(
np
.
uint8
))
elif
args
.
model_net
==
'StarGAN'
:
elif
args
.
model_net
==
'StarGAN'
:
...
@@ -187,7 +189,7 @@ def infer(args):
...
@@ -187,7 +189,7 @@ def infer(args):
tensor_label_org
=
fluid
.
LoDTensor
()
tensor_label_org
=
fluid
.
LoDTensor
()
tensor_img
.
set
(
real_img
,
place
)
tensor_img
.
set
(
real_img
,
place
)
tensor_label_org
.
set
(
label_org
,
place
)
tensor_label_org
.
set
(
label_org
,
place
)
real_img_temp
=
np
.
squeeze
(
real_img
).
transpose
([
0
,
2
,
3
,
1
]
)
real_img_temp
=
save_batch_image
(
real_img
)
images
=
[
real_img_temp
]
images
=
[
real_img_temp
]
for
i
in
range
(
args
.
c_dim
):
for
i
in
range
(
args
.
c_dim
):
label_trg_tmp
=
copy
.
deepcopy
(
label_org
)
label_trg_tmp
=
copy
.
deepcopy
(
label_org
)
...
@@ -201,10 +203,11 @@ def infer(args):
...
@@ -201,10 +203,11 @@ def infer(args):
feed
=
{
"input"
:
tensor_img
,
feed
=
{
"input"
:
tensor_img
,
"label_trg_"
:
tensor_label_trg
},
"label_trg_"
:
tensor_label_trg
},
fetch_list
=
[
fake
.
name
])
fetch_list
=
[
fake
.
name
])
fake_temp
=
np
.
squeeze
(
out
[
0
]).
transpose
([
0
,
2
,
3
,
1
])
fake_temp
=
save_batch_image
(
out
[
0
])
images
.
append
(
fake_temp
)
images
.
append
(
fake_temp
)
images_concat
=
np
.
concatenate
(
images
,
1
)
images_concat
=
np
.
concatenate
(
images
,
1
)
images_concat
=
np
.
concatenate
(
images_concat
,
1
)
if
len
(
label_org
)
>
1
:
images_concat
=
np
.
concatenate
(
images_concat
,
1
)
imageio
.
imwrite
(
args
.
output
+
"/fake_img_"
+
name
[
0
],
(
imageio
.
imwrite
(
args
.
output
+
"/fake_img_"
+
name
[
0
],
(
(
images_concat
+
1
)
*
127.5
).
astype
(
np
.
uint8
))
(
images_concat
+
1
)
*
127.5
).
astype
(
np
.
uint8
))
...
...
PaddleCV/PaddleGAN/network/AttGAN_network.py
浏览文件 @
23d450e9
...
@@ -55,6 +55,7 @@ class AttGAN_model(object):
...
@@ -55,6 +55,7 @@ class AttGAN_model(object):
name
=
name
,
name
=
name
,
dim
=
cfg
.
d_base_dims
,
dim
=
cfg
.
d_base_dims
,
fc_dim
=
cfg
.
d_fc_dim
,
fc_dim
=
cfg
.
d_fc_dim
,
norm
=
cfg
.
dis_norm
,
n_layers
=
cfg
.
n_layers
)
n_layers
=
cfg
.
n_layers
)
def
concat
(
self
,
z
,
a
):
def
concat
(
self
,
z
,
a
):
...
@@ -149,11 +150,11 @@ class AttGAN_model(object):
...
@@ -149,11 +150,11 @@ class AttGAN_model(object):
d
,
d
,
4
,
4
,
2
,
2
,
norm
=
None
,
norm
=
norm
,
padding
=
1
,
padding
=
1
,
activation_fn
=
'leaky_relu'
,
activation_fn
=
'leaky_relu'
,
name
=
name
+
str
(
i
),
name
=
name
+
str
(
i
),
use_bias
=
True
,
use_bias
=
(
norm
==
None
)
,
relufactor
=
0.01
,
relufactor
=
0.01
,
initial
=
'kaiming'
)
initial
=
'kaiming'
)
...
...
PaddleCV/PaddleGAN/network/CGAN_network.py
浏览文件 @
23d450e9
...
@@ -35,6 +35,10 @@ class CGAN_model(object):
...
@@ -35,6 +35,10 @@ class CGAN_model(object):
self
.
gf_dim
=
128
self
.
gf_dim
=
128
self
.
df_dim
=
64
self
.
df_dim
=
64
self
.
leaky_relu_factor
=
0.2
self
.
leaky_relu_factor
=
0.2
if
self
.
batch_size
==
1
:
self
.
norm
=
None
else
:
self
.
norm
=
"batch_norm"
def
network_G
(
self
,
input
,
label
,
name
=
"generator"
):
def
network_G
(
self
,
input
,
label
,
name
=
"generator"
):
# concat noise and label
# concat noise and label
...
@@ -43,14 +47,14 @@ class CGAN_model(object):
...
@@ -43,14 +47,14 @@ class CGAN_model(object):
o_l1
=
linear
(
o_l1
=
linear
(
xy
,
xy
,
self
.
gf_dim
*
8
,
self
.
gf_dim
*
8
,
norm
=
'batch_norm'
,
norm
=
self
.
norm
,
activation_fn
=
'relu'
,
activation_fn
=
'relu'
,
name
=
name
+
'_l1'
)
name
=
name
+
'_l1'
)
o_c1
=
fluid
.
layers
.
concat
([
o_l1
,
y
],
1
)
o_c1
=
fluid
.
layers
.
concat
([
o_l1
,
y
],
1
)
o_l2
=
linear
(
o_l2
=
linear
(
o_c1
,
o_c1
,
self
.
gf_dim
*
(
self
.
img_w
//
4
)
*
(
self
.
img_h
//
4
),
self
.
gf_dim
*
(
self
.
img_w
//
4
)
*
(
self
.
img_h
//
4
),
norm
=
'batch_norm'
,
norm
=
self
.
norm
,
activation_fn
=
'relu'
,
activation_fn
=
'relu'
,
name
=
name
+
'_l2'
)
name
=
name
+
'_l2'
)
o_r1
=
fluid
.
layers
.
reshape
(
o_r1
=
fluid
.
layers
.
reshape
(
...
@@ -107,7 +111,7 @@ class CGAN_model(object):
...
@@ -107,7 +111,7 @@ class CGAN_model(object):
o_l3
=
linear
(
o_l3
=
linear
(
o_c2
,
o_c2
,
self
.
df_dim
*
16
,
self
.
df_dim
*
16
,
norm
=
'batch_norm'
,
norm
=
self
.
norm
,
activation_fn
=
'leaky_relu'
,
activation_fn
=
'leaky_relu'
,
name
=
name
+
'_l3'
)
name
=
name
+
'_l3'
)
o_c3
=
fluid
.
layers
.
concat
([
o_l3
,
y
],
1
)
o_c3
=
fluid
.
layers
.
concat
([
o_l3
,
y
],
1
)
...
...
PaddleCV/PaddleGAN/network/DCGAN_network.py
浏览文件 @
23d450e9
...
@@ -31,13 +31,17 @@ class DCGAN_model(object):
...
@@ -31,13 +31,17 @@ class DCGAN_model(object):
self
.
dfc_dim
=
1024
self
.
dfc_dim
=
1024
self
.
gf_dim
=
64
self
.
gf_dim
=
64
self
.
df_dim
=
64
self
.
df_dim
=
64
if
self
.
batch_size
==
1
:
self
.
norm
=
None
else
:
self
.
norm
=
"batch_norm"
def
network_G
(
self
,
input
,
name
=
"generator"
):
def
network_G
(
self
,
input
,
name
=
"generator"
):
o_l1
=
linear
(
input
,
self
.
gfc_dim
,
norm
=
'batch_norm'
,
name
=
name
+
'_l1'
)
o_l1
=
linear
(
input
,
self
.
gfc_dim
,
norm
=
self
.
norm
,
name
=
name
+
'_l1'
)
o_l2
=
linear
(
o_l2
=
linear
(
o_l1
,
o_l1
,
self
.
gf_dim
*
2
*
self
.
img_dim
//
4
*
self
.
img_dim
//
4
,
self
.
gf_dim
*
2
*
self
.
img_dim
//
4
*
self
.
img_dim
//
4
,
norm
=
'batch_norm'
,
norm
=
self
.
norm
,
name
=
name
+
'_l2'
)
name
=
name
+
'_l2'
)
o_r1
=
fluid
.
layers
.
reshape
(
o_r1
=
fluid
.
layers
.
reshape
(
o_l2
,
[
-
1
,
self
.
df_dim
*
2
,
self
.
img_dim
//
4
,
self
.
img_dim
//
4
])
o_l2
,
[
-
1
,
self
.
df_dim
*
2
,
self
.
img_dim
//
4
,
self
.
img_dim
//
4
])
...
@@ -85,7 +89,7 @@ class DCGAN_model(object):
...
@@ -85,7 +89,7 @@ class DCGAN_model(object):
o_l1
=
linear
(
o_l1
=
linear
(
o_c2
,
o_c2
,
self
.
dfc_dim
,
self
.
dfc_dim
,
norm
=
'batch_norm'
,
norm
=
self
.
norm
,
activation_fn
=
'leaky_relu'
,
activation_fn
=
'leaky_relu'
,
name
=
name
+
'_l1'
)
name
=
name
+
'_l1'
)
out
=
linear
(
o_l1
,
1
,
activation_fn
=
'sigmoid'
,
name
=
name
+
'_l2'
)
out
=
linear
(
o_l1
,
1
,
activation_fn
=
'sigmoid'
,
name
=
name
+
'_l2'
)
...
...
PaddleCV/PaddleGAN/network/STGAN_network.py
浏览文件 @
23d450e9
...
@@ -76,6 +76,7 @@ class STGAN_model(object):
...
@@ -76,6 +76,7 @@ class STGAN_model(object):
n_atts
=
cfg
.
c_dim
,
n_atts
=
cfg
.
c_dim
,
dim
=
cfg
.
d_base_dims
,
dim
=
cfg
.
d_base_dims
,
fc_dim
=
cfg
.
d_fc_dim
,
fc_dim
=
cfg
.
d_fc_dim
,
norm
=
cfg
.
dis_norm
,
n_layers
=
cfg
.
n_layers
,
n_layers
=
cfg
.
n_layers
,
name
=
name
)
name
=
name
)
...
@@ -100,7 +101,7 @@ class STGAN_model(object):
...
@@ -100,7 +101,7 @@ class STGAN_model(object):
activation_fn
=
'leaky_relu'
,
activation_fn
=
'leaky_relu'
,
name
=
name
+
str
(
i
),
name
=
name
+
str
(
i
),
use_bias
=
False
,
use_bias
=
False
,
relufactor
=
0.
01
,
relufactor
=
0.
2
,
initial
=
'kaiming'
,
initial
=
'kaiming'
,
is_test
=
is_test
)
is_test
=
is_test
)
zs
.
append
(
z
)
zs
.
append
(
z
)
...
@@ -132,7 +133,7 @@ class STGAN_model(object):
...
@@ -132,7 +133,7 @@ class STGAN_model(object):
pass_state
=
pass_state
,
pass_state
=
pass_state
,
name
=
name
+
str
(
i
),
name
=
name
+
str
(
i
),
is_test
=
is_test
)
is_test
=
is_test
)
zs_
.
insert
(
0
,
output
[
0
]
+
zs
[
n_layers
-
1
-
i
]
)
zs_
.
insert
(
0
,
output
[
0
])
if
inject_layers
>
i
:
if
inject_layers
>
i
:
state
=
self
.
concat
(
output
[
1
],
a
)
state
=
self
.
concat
(
output
[
1
],
a
)
else
:
else
:
...
@@ -202,18 +203,18 @@ class STGAN_model(object):
...
@@ -202,18 +203,18 @@ class STGAN_model(object):
d
,
d
,
4
,
4
,
2
,
2
,
norm
=
None
,
norm
=
norm
,
padding
=
1
,
padding
_type
=
"SAME"
,
activation_fn
=
'leaky_relu'
,
activation_fn
=
'leaky_relu'
,
name
=
name
+
str
(
i
),
name
=
name
+
str
(
i
),
use_bias
=
True
,
use_bias
=
(
norm
==
None
)
,
relufactor
=
0.
01
,
relufactor
=
0.
2
,
initial
=
'kaiming'
)
initial
=
'kaiming'
)
logit_gan
=
linear
(
logit_gan
=
linear
(
y
,
y
,
fc_dim
,
fc_dim
,
activation_fn
=
'relu'
,
activation_fn
=
'
leaky_
relu'
,
name
=
name
+
'fc_adv_1'
,
name
=
name
+
'fc_adv_1'
,
initial
=
'kaiming'
)
initial
=
'kaiming'
)
logit_gan
=
linear
(
logit_gan
=
linear
(
...
@@ -222,7 +223,7 @@ class STGAN_model(object):
...
@@ -222,7 +223,7 @@ class STGAN_model(object):
logit_att
=
linear
(
logit_att
=
linear
(
y
,
y
,
fc_dim
,
fc_dim
,
activation_fn
=
'relu'
,
activation_fn
=
'
leaky_
relu'
,
name
=
name
+
'fc_cls_1'
,
name
=
name
+
'fc_cls_1'
,
initial
=
'kaiming'
)
initial
=
'kaiming'
)
logit_att
=
linear
(
logit_att
=
linear
(
...
...
PaddleCV/PaddleGAN/network/base_network.py
浏览文件 @
23d450e9
...
@@ -76,7 +76,7 @@ def norm_layer(input, norm_type='batch_norm', name=None, is_test=False):
...
@@ -76,7 +76,7 @@ def norm_layer(input, norm_type='batch_norm', name=None, is_test=False):
tmp
=
fluid
.
layers
.
elementwise_add
(
tmp
,
offset
,
axis
=
1
)
tmp
=
fluid
.
layers
.
elementwise_add
(
tmp
,
offset
,
axis
=
1
)
return
tmp
return
tmp
else
:
else
:
raise
NotImplementedError
(
"norm ty
o
e: [%s] is not support"
%
norm_type
)
raise
NotImplementedError
(
"norm ty
p
e: [%s] is not support"
%
norm_type
)
def
initial_type
(
name
,
def
initial_type
(
name
,
...
...
PaddleCV/PaddleGAN/scripts/run_attgan.sh
浏览文件 @
23d450e9
python train.py
--model_net
AttGAN
--dataset
celeba
--crop_size
170
--image_size
128
--train_list
./data/celeba/list_attr_celeba.txt
--test_list
./data/celeba/test_list_attr_celeba.txt
--gan_mode
wgan
--batch_size
32
--print_freq
1
--num_discriminator_time
5
--epoch
90
>
log
.
out 2>log_err
python train.py
--model_net
AttGAN
--dataset
celeba
--crop_size
170
--image_size
128
--train_list
./data/celeba/list_attr_celeba.txt
--test_list
./data/celeba/test_list_attr_celeba.txt
--gan_mode
wgan
--batch_size
32
--print_freq
1
--num_discriminator_time
5
--epoch
90
>
log
_
out 2>log_err
PaddleCV/PaddleGAN/scripts/run_stgan.sh
浏览文件 @
23d450e9
python train.py
--model_net
STGAN
--dataset
celeba
--crop_size
170
--image_size
128
--train_list
./data/celeba/list_attr_celeba.txt
--test_list
./data/celeba/test_list_attr_celeba.txt
--gan_mode
wgan
--batch_size
32
--print_freq
1
--num_discriminator_time
5
--epoch
20
>
log.
out 2>log_err
python train.py
--model_net
STGAN
--dataset
celeba
--crop_size
170
--image_size
128
--train_list
./data/celeba/list_attr_celeba.txt
--test_list
./data/celeba/test_list_attr_celeba.txt
--gan_mode
wgan
--batch_size
32
--print_freq
1
--num_discriminator_time
5
--epoch
50
>
log_
out 2>log_err
PaddleCV/PaddleGAN/trainer/AttGAN.py
浏览文件 @
23d450e9
...
@@ -229,6 +229,12 @@ class AttGAN(object):
...
@@ -229,6 +229,12 @@ class AttGAN(object):
type
=
int
,
type
=
int
,
default
=
5
,
default
=
5
,
help
=
"default layers in the network"
)
help
=
"default layers in the network"
)
parser
.
add_argument
(
'--dis_norm'
,
type
=
str
,
default
=
None
,
help
=
"the normalization in discriminator, choose in [None, instance_norm]"
)
return
parser
return
parser
...
...
PaddleCV/PaddleGAN/trainer/STGAN.py
浏览文件 @
23d450e9
...
@@ -79,7 +79,7 @@ class DTrainer():
...
@@ -79,7 +79,7 @@ class DTrainer():
clone_image_real
=
b
.
var
(
'image_real'
)
clone_image_real
=
b
.
var
(
'image_real'
)
break
break
self
.
fake_img
,
_
=
model
.
network_G
(
self
.
fake_img
,
_
=
model
.
network_G
(
image_real
,
label_org
,
label_trg_
,
cfg
,
name
=
"generator"
)
image_real
,
label_org
_
,
label_trg_
,
cfg
,
name
=
"generator"
)
self
.
pred_real
,
self
.
cls_real
=
model
.
network_D
(
self
.
pred_real
,
self
.
cls_real
=
model
.
network_D
(
image_real
,
cfg
,
name
=
"discriminator"
)
image_real
,
cfg
,
name
=
"discriminator"
)
self
.
pred_real
.
persistable
=
True
self
.
pred_real
.
persistable
=
True
...
@@ -234,6 +234,12 @@ class STGAN(object):
...
@@ -234,6 +234,12 @@ class STGAN(object):
type
=
int
,
type
=
int
,
default
=
4
,
default
=
4
,
help
=
"default layers of GRU in generotor"
)
help
=
"default layers of GRU in generotor"
)
parser
.
add_argument
(
'--dis_norm'
,
type
=
str
,
default
=
None
,
help
=
"the normalization in discriminator, choose in [None, instance_norm]"
)
return
parser
return
parser
...
...
PaddleCV/PaddleGAN/util/utility.py
浏览文件 @
23d450e9
...
@@ -109,13 +109,13 @@ def save_test_image(epoch,
...
@@ -109,13 +109,13 @@ def save_test_image(epoch,
tensor_label_org
=
fluid
.
LoDTensor
()
tensor_label_org
=
fluid
.
LoDTensor
()
tensor_img
.
set
(
real_img
,
place
)
tensor_img
.
set
(
real_img
,
place
)
tensor_label_org
.
set
(
label_org
,
place
)
tensor_label_org
.
set
(
label_org
,
place
)
real_img_temp
=
np
.
squeeze
(
real_img
).
transpose
([
1
,
2
,
0
]
)
real_img_temp
=
save_batch_image
(
real_img
)
images
=
[
real_img_temp
]
images
=
[
real_img_temp
]
for
i
in
range
(
cfg
.
c_dim
):
for
i
in
range
(
cfg
.
c_dim
):
label_trg_tmp
=
copy
.
deepcopy
(
label_org
)
label_trg_tmp
=
copy
.
deepcopy
(
label_org
)
label_trg_tmp
[
0
][
i
]
=
1.0
-
label_trg_tmp
[
0
][
i
]
label_trg_tmp
[
0
][
i
]
=
1.0
-
label_trg_tmp
[
0
][
i
]
label_trg
=
check_attribute_conflict
(
label_trg
=
check_attribute_conflict
(
label_trg_tmp
,
label_trg_tmp
,
attr_names
[
i
],
attr_names
)
attr_names
[
i
],
attr_names
)
tensor_label_trg
=
fluid
.
LoDTensor
()
tensor_label_trg
=
fluid
.
LoDTensor
()
tensor_label_trg
.
set
(
label_trg
,
place
)
tensor_label_trg
.
set
(
label_trg
,
place
)
fake_temp
,
rec_temp
=
exe
.
run
(
fake_temp
,
rec_temp
=
exe
.
run
(
...
@@ -126,8 +126,8 @@ def save_test_image(epoch,
...
@@ -126,8 +126,8 @@ def save_test_image(epoch,
"label_trg"
:
tensor_label_trg
"label_trg"
:
tensor_label_trg
},
},
fetch_list
=
[
g_trainer
.
fake_img
,
g_trainer
.
rec_img
])
fetch_list
=
[
g_trainer
.
fake_img
,
g_trainer
.
rec_img
])
fake_temp
=
np
.
squeeze
(
fake_temp
[
0
]).
transpose
([
1
,
2
,
0
])
fake_temp
=
save_batch_image
(
fake_temp
[
0
])
rec_temp
=
np
.
squeeze
(
rec_temp
[
0
]).
transpose
([
1
,
2
,
0
])
rec_temp
=
save_batch_image
(
rec_temp
[
0
])
images
.
append
(
fake_temp
)
images
.
append
(
fake_temp
)
images
.
append
(
rec_temp
)
images
.
append
(
rec_temp
)
images_concat
=
np
.
concatenate
(
images
,
1
)
images_concat
=
np
.
concatenate
(
images
,
1
)
...
@@ -145,7 +145,7 @@ def save_test_image(epoch,
...
@@ -145,7 +145,7 @@ def save_test_image(epoch,
tensor_label_trg_
=
fluid
.
LoDTensor
()
tensor_label_trg_
=
fluid
.
LoDTensor
()
tensor_img
.
set
(
real_img
,
place
)
tensor_img
.
set
(
real_img
,
place
)
tensor_label_org
.
set
(
label_org
,
place
)
tensor_label_org
.
set
(
label_org
,
place
)
real_img_temp
=
np
.
squeeze
(
real_img
).
transpose
([
0
,
2
,
3
,
1
]
)
real_img_temp
=
save_batch_image
(
real_img
)
images
=
[
real_img_temp
]
images
=
[
real_img_temp
]
for
i
in
range
(
cfg
.
c_dim
):
for
i
in
range
(
cfg
.
c_dim
):
label_trg_tmp
=
copy
.
deepcopy
(
label_trg
)
label_trg_tmp
=
copy
.
deepcopy
(
label_trg
)
...
@@ -155,12 +155,14 @@ def save_test_image(epoch,
...
@@ -155,12 +155,14 @@ def save_test_image(epoch,
label_trg_tmp
=
check_attribute_conflict
(
label_trg_tmp
=
check_attribute_conflict
(
label_trg_tmp
,
attr_names
[
i
],
attr_names
)
label_trg_tmp
,
attr_names
[
i
],
attr_names
)
label_org_
=
list
(
map
(
lambda
x
:
((
x
*
2
)
-
1
)
*
0.5
,
label_org
))
label_trg_
=
list
(
label_trg_
=
list
(
map
(
lambda
x
:
((
x
*
2
)
-
1
)
*
0.5
,
label_trg_tmp
))
map
(
lambda
x
:
((
x
*
2
)
-
1
)
*
0.5
,
label_trg_tmp
))
for
j
in
range
(
len
(
label_org
)):
if
cfg
.
model_net
==
'AttGAN'
:
label_trg_
[
j
][
i
]
=
label_trg_
[
j
][
i
]
*
2.0
for
k
in
range
(
len
(
label_org
)):
tensor_label_org_
.
set
(
label_org
,
place
)
label_trg_
[
k
][
i
]
=
label_trg_
[
k
][
i
]
*
2.0
tensor_label_org_
.
set
(
label_org_
,
place
)
tensor_label_trg
.
set
(
label_trg
,
place
)
tensor_label_trg
.
set
(
label_trg
,
place
)
tensor_label_trg_
.
set
(
label_trg_
,
place
)
tensor_label_trg_
.
set
(
label_trg_
,
place
)
out
=
exe
.
run
(
test_program
,
out
=
exe
.
run
(
test_program
,
...
@@ -172,10 +174,11 @@ def save_test_image(epoch,
...
@@ -172,10 +174,11 @@ def save_test_image(epoch,
"label_trg_"
:
tensor_label_trg_
"label_trg_"
:
tensor_label_trg_
},
},
fetch_list
=
[
g_trainer
.
fake_img
])
fetch_list
=
[
g_trainer
.
fake_img
])
fake_temp
=
np
.
squeeze
(
out
[
0
]).
transpose
([
0
,
2
,
3
,
1
])
fake_temp
=
save_batch_image
(
out
[
0
])
images
.
append
(
fake_temp
)
images
.
append
(
fake_temp
)
images_concat
=
np
.
concatenate
(
images
,
1
)
images_concat
=
np
.
concatenate
(
images
,
1
)
images_concat
=
np
.
concatenate
(
images_concat
,
1
)
if
len
(
label_org
)
>
1
:
images_concat
=
np
.
concatenate
(
images_concat
,
1
)
imageio
.
imwrite
(
out_path
+
"/fake_img"
+
str
(
epoch
)
+
'_'
+
name
[
0
],
imageio
.
imwrite
(
out_path
+
"/fake_img"
+
str
(
epoch
)
+
'_'
+
name
[
0
],
((
images_concat
+
1
)
*
127.5
).
astype
(
np
.
uint8
))
((
images_concat
+
1
)
*
127.5
).
astype
(
np
.
uint8
))
...
@@ -266,20 +269,28 @@ def check_attribute_conflict(label_batch, attr, attrs):
...
@@ -266,20 +269,28 @@ def check_attribute_conflict(label_batch, attr, attrs):
return
label_batch
return
label_batch
def
save_batch_image
(
img
):
if
len
(
img
)
==
1
:
res_img
=
np
.
squeeze
(
img
).
transpose
([
1
,
2
,
0
])
else
:
res_img
=
np
.
squeeze
(
img
).
transpose
([
0
,
2
,
3
,
1
])
return
res_img
def
check_gpu
(
use_gpu
):
def
check_gpu
(
use_gpu
):
"""
"""
Log error and exit when set use_gpu=true in paddlepaddle
Log error and exit when set use_gpu=true in paddlepaddle
cpu version.
cpu version.
"""
"""
err
=
"Config use_gpu cannot be set as true while you are "
\
err
=
"Config use_gpu cannot be set as true while you are "
\
"using paddlepaddle cpu version !
\n
Please try:
\n
"
\
"using paddlepaddle cpu version !
\n
Please try:
\n
"
\
"
\t
1. Install paddlepaddle-gpu to run model on GPU
\n
"
\
"
\t
1. Install paddlepaddle-gpu to run model on GPU
\n
"
\
"
\t
2. Set use_gpu as false in config file to run "
\
"
\t
2. Set use_gpu as false in config file to run "
\
"model on CPU"
"model on CPU"
try
:
try
:
if
use_gpu
and
not
fluid
.
is_compiled_with_cuda
():
if
use_gpu
and
not
fluid
.
is_compiled_with_cuda
():
logger
.
error
(
err
)
logger
.
error
(
err
)
sys
.
exit
(
1
)
sys
.
exit
(
1
)
except
Exception
as
e
:
except
Exception
as
e
:
pass
pass
编辑
预览
Markdown
is supported
0%
请重试
或
添加新附件
.
添加附件
取消
You are about to add
0
people
to the discussion. Proceed with caution.
先完成此消息的编辑!
取消
想要评论请
注册
或
登录