CycleGAN（一）概览与运行

背景：

源码作者：junyanz/pytorch-CycleGAN-and-pix2pix

源码地址：https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix

参考：https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix

目录

一、下载

1.1 下载

1.2 安装好依赖项

二、训练与测试

2.1 下载数据集

2.2. 自备数据集

2.3 可视化

2.4 训练

2.5 测试

三、预训练模型

3.1 预训练模型

3.2 数据集

四、docker

---

一、下载

1.1 下载

git clone https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix
cd pytorch-CycleGAN-and-pix2pix

1.2 安装好依赖项

PyTorch安装，见：https://blog.csdn.net/weixin_36474809/article/details/88715360

Install [PyTorch](http://pytorch.org and) 0.4+ and other dependencies (e.g., torchvision, visdom and dominate).

• For pip users, please type the command pip install -r requirements.txt

但是对于我们而言，因为之前装好的python在env/bin/python，所以我们需要在上级文件夹输入

env/bin/python -m pip install -r pytorch-CycleGAN-and-pix2pix/requirements.txt

二、训练与测试

2.1 下载数据集

我们以map数据集为例子。

对我们而言，出现问题

[xingxiangrui@xxxxxxxx pytorch-CycleGAN-and-pix2pix]$ bash ./datasets/download_cyclegan_dataset.sh maps
Specified [maps]
WARNING: timestamping does nothing in combination with -O. See the manual
for details.

--2019-03-22 11:07:20--  https://people.eecs.berkeley.edu/~taesung_park/CycleGAN/datasets/maps.zip
Resolving people.eecs.berkeley.edu... 128.32.189.73
Connecting to people.eecs.berkeley.edu|128.32.189.73|:443... connected.
ERROR: certificate common name "iris.eecs.berkeley.edu" doesn't match requested host name "people.eecs.berkeley.edu".
To connect to people.eecs.berkeley.edu insecurely, use '--no-check-certificate'.
mkdir: cannot create directory './datasets/maps/': File exists
Archive:  ./datasets/maps.zip
  End-of-central-directory signature not found.  Either this file is not
  a zipfile, or it constitutes one disk of a multi-part archive.  In the
  latter case the central directory and zipfile comment will be found on
  the last disk(s) of this archive.
unzip:  cannot find zipfile directory in one of ./datasets/maps.zip or
        ./datasets/maps.zip.zip, and cannot find ./datasets/maps.zip.ZIP, period.

其他所有数据集均无法下载：例如下面网页之中apple2orange和horse2zebra等等

https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix/blob/master/docs/datasets.md

数据集的下载出现问题，所以我们只好自备数据集。

2.2. 自备数据集

作者有给出一个自制模型和数据集的方法：

准备一个文件夹try，里面包含四个文件夹trainA、trainB、testA、testB，每个文件夹内分别放有若干图片

复制到 pytorch-CycleGAN-and-pix2pix/datasets/ 下

https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix/blob/master/docs/tips.md#notes-on-colorization

2.3 可视化

可视化用的是visdom.server

To view training results and loss plots, run python -m visdom.server and click the URL http://localhost:8097

关于visdom.server的用法我们暂时还没确定清楚

2.4 训练

#!./scripts/train_cyclegan.sh
python train.py --dataroot ./datasets/maps --name maps_cyclegan --model cycle_gan 

可以看出其中

• --dataroot 训练数据的根目录
• --name 训练模型的名称
• --model 训练的模型的类型

更多训练时的信息查看 ./checkpoints/maps_cyclegan/web/index.html

2.5 测试

#!./scripts/test_cyclegan.sh
python test.py --dataroot ./datasets/maps --name maps_cyclegan --model cycle_gan

测试结果存于 ./results/maps_cyclegan/latest_test/index.html

三、预训练模型

3.1 预训练模型

• You can download a pretrained model (e.g. horse2zebra) with the following script:

bash ./scripts/download_cyclegan_model.sh horse2zebra

• The pretrained model is saved at ./checkpoints/{name}_pretrained/latest_net_G.pth. Check here for all the available CycleGAN models.

3.2 数据集

• To test the model, you also need to download the horse2zebra dataset:

bash ./datasets/download_cyclegan_dataset.sh horse2zebra

• Then generate the results using

python test.py --dataroot datasets/horse2zebra/testA --name horse2zebra_pretrained --model test --no_dropout

• The option --model test is used for generating results of CycleGAN only for one side. This option will automatically set --dataset_mode single, which only loads the images from one set. On the contrary, using --model cycle_gan requires loading and generating results in both directions, which is sometimes unnecessary. The results will be saved at ./results/. Use --results_dir {directory_path_to_save_result} to specify the results directory.

• For your own experiments, you might want to specify --netG, --norm, --no_dropout to match the generator architecture of the trained model.

四、docker

https://baike.baidu.com/item/Docker/13344470?fr=aladdin

Docker 是一个开源的应用容器引擎，让开发者可以打包他们的应用以及依赖包到一个可移植的容器中，然后发布到任何流行的 Linux 机器上，也可以实现虚拟化。容器是完全使用沙箱机制，相互之间不会有任何接口。

https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix/blob/master/docs/docker.md