CycleGAN（二）数据集重做与训练测试

阳光穿透心脏的1/2处 2022-03-01 12:54 1520阅读 0赞

目的：我们需要在我们的数据集上实现CycleGAN。

参考：[https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix/blob/master/docs/datasets.md][https_github.com_junyanz_pytorch-CycleGAN-and-pix2pix_blob_master_docs_datasets.md]

**目录**

一、数据集制作

1.1 数据集格式

1.2 我们的数据集

二、训练

1.1 命令行

1.2 不显示结果

二、训练命令行

2.1 norText\_2\_cotton

2.2 norText\_2\_falText

三、测试

3.1 测试相关命令行及含义

两点注意

3.2 测试

四、存储结果位置

4.1 测试结果位置

html

4.2 训练结果位置

五、loss常见值

5.1 norText\_2\_falText\_cyclegan

开始

--------------------

# 一、数据集制作 #

## 1.1 数据集格式 ##

To train a model on your own datasets, you need to create a data folder with two subdirectories `trainA` and `trainB` that contain images from domain A and B. You can test your model on your training set by setting `--phase train` in `test.py`. You can also create subdirectories `testA` and `testB` if you have test data.

*  创建文件夹，子文件夹中包含了trainA，trainB，分别是A域和B域。
 *  测试时，在test.py中设置--phase train从而对模型进行test。
 *  然后创建子文件夹testA和testB。

注意，trainA和trainB需要相似的图像，例如：`cats<->keyboards这种可能不work，因为domaionA与domainB相差太远，但是landscape painting<->landscape photographs这种就很可能work，因为两域差别并不大。`

## 1.2 我们的数据集 ##

创建norText\_2\_contton数据集

其中四个文件夹 trainA是100张norText数据，trainB为24张cotton数据，testA为norText种抽出的20张正常数据，testB为空，到时生成的数据就在testB之中。

本地压缩，然后上传到服务器datasets之下解压

rz
    # Received /Users/baidu/Desktop/实习文档/纺织品数据集/norText_2_cotton.zip
    ls
    unzip norText_2_cotton.zip

# 二、训练 #

## 1.1 命令行 ##

命令行参考:

python train.py --dataroot ./datasets/maps --name maps_cyclegan --model cycle_gan

*  我们的训练集的路径：--dataroot  /home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/datasets/norText\_2\_cotton
 *  训练好的模型的名称为 --name norText\_2\_cotton\_cyclegan
 *  模型类型为 --model cycle\_gan

所以最开始输入的命令行为：

env/bin/python /home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/train.py --dataroot /home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/datasets/norText\_2\_cotton --name norText\_2\_cotton\_cyclegan --model cycle\_gan

后续不显示结果需要在后面加`--no_html，我们命令行以此为准:`

运行成功：

[xingxiangrui@8888888888888888 ~]$ env/bin/python /home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/train.py --dataroot /home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/datasets/norText_2_cotton --name norText_2_cotton_cyclegan --model cycle_gan
    ----------------- Options ---------------
                   batch_size: 1
                        beta1: 0.5
              checkpoints_dir: ./checkpoints
               continue_train: False
                    crop_size: 256
                     dataroot: /home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/datasets/norText_2_cotton	[default: None]
                 dataset_mode: unaligned
                    direction: AtoB
                  display_env: main
                 display_freq: 400
                   display_id: 1
                display_ncols: 4
                 display_port: 8097
               display_server: http://localhost
              display_winsize: 256
                        epoch: latest
                  epoch_count: 1
                     gan_mode: lsgan
                      gpu_ids: 0
                    init_gain: 0.02
                    init_type: normal
                     input_nc: 3
                      isTrain: True                          	[default: None]
                     lambda_A: 10.0
                     lambda_B: 10.0
              lambda_identity: 0.5
                    load_iter: 0                             	[default: 0]
                    load_size: 286
                           lr: 0.0002
               lr_decay_iters: 50
                    lr_policy: linear
             max_dataset_size: inf
                        model: cycle_gan
                   n_layers_D: 3
                         name: norText_2_cotton_cyclegan     	[default: experiment_name]
                          ndf: 64
                         netD: basic
                         netG: resnet_9blocks
                          ngf: 64
                        niter: 100
                  niter_decay: 100
                   no_dropout: True
                      no_flip: False
                      no_html: False
                         norm: instance
                  num_threads: 4
                    output_nc: 3
                        phase: train
                    pool_size: 50
                   preprocess: resize_and_crop
                   print_freq: 100
                 save_by_iter: False
              save_epoch_freq: 5
             save_latest_freq: 5000
               serial_batches: False
                       suffix:
             update_html_freq: 1000
                      verbose: False
    ----------------- End -------------------
    dataset [UnalignedDataset] was created
    The number of training images = 100
    initialize network with normal
    initialize network with normal
    initialize network with normal
    initialize network with normal
    model [CycleGANModel] was created
    ---------- Networks initialized -------------
    [Network G_A] Total number of parameters : 11.378 M
    [Network G_B] Total number of parameters : 11.378 M
    [Network D_A] Total number of parameters : 2.765 M
    [Network D_B] Total number of parameters : 2.765 M
    -----------------------------------------------
    。。。
    create web directory ./checkpoints/norText_2_cotton_cyclegan/web...
    (epoch: 1, iters: 100, time: 0.896, data: 1.052) D_A: 0.384 G_A: 0.232 cycle_A: 1.791 idt_A: 0.739 D_B: 0.572 G_B: 0.620 cycle_B: 2.002 idt_B: 0.851
    End of epoch 1 / 200 	 Time Taken: 92 sec
    learning rate = 0.0002000
    (epoch: 2, iters: 100, time: 0.865, data: 0.214) D_A: 0.219 G_A: 0.304 cycle_A: 1.499 idt_A: 0.597 D_B: 0.305 G_B: 0.699 cycle_B: 1.118 idt_B: 0.711
    End of epoch 2 / 200 	 Time Taken: 87 sec
    learning rate = 0.0002000
    。。。

看样子模型会迭代200个epoch。最终的loss情况：

(epoch: 195, iters: 100, time: 0.865, data: 0.209) D_A: 0.016 G_A: 0.836 cycle_A: 0.584 idt_A: 0.168 D_B: 0.124 G_B: 0.346 cycle_B: 0.530 idt_B: 0.181
    saving the model at the end of epoch 195, iters 19500
    End of epoch 195 / 200 	 Time Taken: 88 sec
    learning rate = 0.0000119
    (epoch: 196, iters: 100, time: 1.197, data: 0.237) D_A: 0.036 G_A: 0.737 cycle_A: 0.590 idt_A: 0.133 D_B: 0.014 G_B: 0.271 cycle_B: 0.425 idt_B: 0.186
    End of epoch 196 / 200 	 Time Taken: 87 sec
    learning rate = 0.0000099
    (epoch: 197, iters: 100, time: 0.871, data: 0.218) D_A: 0.031 G_A: 0.756 cycle_A: 0.533 idt_A: 0.113 D_B: 0.037 G_B: 0.511 cycle_B: 0.370 idt_B: 0.160
    End of epoch 197 / 200 	 Time Taken: 86 sec
    learning rate = 0.0000079
    (epoch: 198, iters: 100, time: 0.856, data: 0.217) D_A: 0.063 G_A: 0.509 cycle_A: 0.634 idt_A: 0.123 D_B: 0.308 G_B: 0.492 cycle_B: 0.478 idt_B: 0.222
    End of epoch 198 / 200 	 Time Taken: 86 sec
    learning rate = 0.0000059
    (epoch: 199, iters: 100, time: 0.903, data: 0.203) D_A: 0.033 G_A: 0.981 cycle_A: 0.515 idt_A: 0.110 D_B: 0.111 G_B: 0.531 cycle_B: 0.381 idt_B: 0.167
    End of epoch 199 / 200 	 Time Taken: 86 sec
    learning rate = 0.0000040
    (epoch: 200, iters: 100, time: 1.200, data: 0.219) D_A: 0.017 G_A: 1.035 cycle_A: 0.613 idt_A: 0.106 D_B: 0.030 G_B: 0.726 cycle_B: 0.384 idt_B: 0.200
    saving the latest model (epoch 200, total_iters 20000)
    saving the model at the end of epoch 200, iters 20000
    End of epoch 200 / 200 	 Time Taken: 89 sec
    learning rate = 0.0000020

## 1.2 不显示结果 ##

模型可以进行可视化中间结果。

During training, the current results can be viewed using two methods. First, if you set `--display_id` > 0, the results and loss plot will appear on a local graphics web server launched by [visdom][]. To do this, you should have `visdom` installed and a server running by the command `python -m visdom.server`. The default server URL is `http://localhost:8097`. `display_id` corresponds to the window ID that is displayed on the `visdom` server. The `visdom` display functionality is turned on by default. To avoid the extra overhead of communicating with `visdom` set `--display_id -1`. Second, the intermediate results are saved to `[opt.checkpoints_dir]/[opt.name]/web/` as an HTML file. To avoid this, set `--no_html`.

我们在运行时因为设置不当，报错。所以我们可以取消此模块的功能，后续再详细研究模型运行中的可视化。

报错信息：

WARNING:root:Setting up a new session...
    Exception in user code:
    ------------------------------------------------------------
    Traceback (most recent call last):
      File "/home/xingxiangrui/env/lib/python3.6/site-packages/urllib3/connection.py", line 159, in _new_conn
        (self._dns_host, self.port), self.timeout, **extra_kw)
      File "/home/xingxiangrui/env/lib/python3.6/site-packages/urllib3/util/connection.py", line 80, in create_connection
        raise err
      File "/home/xingxiangrui/env/lib/python3.6/site-packages/urllib3/util/connection.py", line 70, in create_connection
        sock.connect(sa)
    ConnectionRefusedError: [Errno 111] Connection refused
    
    During handling of the above exception, another exception occurred:
    
    Traceback (most recent call last):
      File "/home/xingxiangrui/env/lib/python3.6/site-packages/urllib3/connectionpool.py", line 600, in urlopen
        chunked=chunked)
      File "/home/xingxiangrui/env/lib/python3.6/site-packages/urllib3/connectionpool.py", line 354, in _make_request
        conn.request(method, url, **httplib_request_kw)
      File "/home/xingxiangrui/env/lib/python3.6/http/client.py", line 1239, in request
        self._send_request(method, url, body, headers, encode_chunked)
      File "/home/xingxiangrui/env/lib/python3.6/http/client.py", line 1285, in _send_request
        self.endheaders(body, encode_chunked=encode_chunked)
      File "/home/xingxiangrui/env/lib/python3.6/http/client.py", line 1234, in endheaders
        self._send_output(message_body, encode_chunked=encode_chunked)
      File "/home/xingxiangrui/env/lib/python3.6/http/client.py", line 1026, in _send_output
        self.send(msg)
      File "/home/xingxiangrui/env/lib/python3.6/http/client.py", line 964, in send
        self.connect()
      File "/home/xingxiangrui/env/lib/python3.6/site-packages/urllib3/connection.py", line 181, in connect
        conn = self._new_conn()
      File "/home/xingxiangrui/env/lib/python3.6/site-packages/urllib3/connection.py", line 168, in _new_conn
        self, "Failed to establish a new connection: %s" % e)
    urllib3.exceptions.NewConnectionError: <urllib3.connection.HTTPConnection object at 0x7fa7f4811b38>: Failed to establish a new connection: [Errno 111] Connection refused
    
    During handling of the above exception, another exception occurred:
    
    Traceback (most recent call last):
      File "/home/xingxiangrui/env/lib/python3.6/site-packages/requests/adapters.py", line 449, in send
        timeout=timeout
      File "/home/xingxiangrui/env/lib/python3.6/site-packages/urllib3/connectionpool.py", line 638, in urlopen
        _stacktrace=sys.exc_info()[2])
      File "/home/xingxiangrui/env/lib/python3.6/site-packages/urllib3/util/retry.py", line 398, in increment
        raise MaxRetryError(_pool, url, error or ResponseError(cause))
    urllib3.exceptions.MaxRetryError: HTTPConnectionPool(host='localhost', port=8097): Max retries exceeded with url: /env/main (Caused by NewConnectionError('<urllib3.connection.HTTPConnection object at 0x7fa7f4811b38>: Failed to establish a new connection: [Errno 111] Connection refused',))
    
    During handling of the above exception, another exception occurred:
    
    Traceback (most recent call last):
      File "/home/xingxiangrui/env/lib/python3.6/site-packages/visdom/__init__.py", line 548, in _send
        data=json.dumps(msg),
      File "/home/xingxiangrui/env/lib/python3.6/site-packages/requests/sessions.py", line 581, in post
        return self.request('POST', url, data=data, json=json, **kwargs)
      File "/home/xingxiangrui/env/lib/python3.6/site-packages/requests/sessions.py", line 533, in request
        resp = self.send(prep, **send_kwargs)
      File "/home/xingxiangrui/env/lib/python3.6/site-packages/requests/sessions.py", line 646, in send
        r = adapter.send(request, **kwargs)
      File "/home/xingxiangrui/env/lib/python3.6/site-packages/requests/adapters.py", line 516, in send
        raise ConnectionError(e, request=request)
    requests.exceptions.ConnectionError: HTTPConnectionPool(host='localhost', port=8097): Max retries exceeded with url: /env/main (Caused by NewConnectionError('<urllib3.connection.HTTPConnection object at 0x7fa7f4811b38>: Failed to establish a new connection: [Errno 111] Connection refused',))
    ERROR:visdom:[Errno 97] Address family not supported by protocol
    ERROR:visdom:[Errno 97] Address family not supported by protocol
    ERROR:visdom:[Errno 97] Address family not supported by protocol
    WARNING:visdom:Visdom python client failed to establish socket to get messages from the server. This feature is optional and can be disabled by initializing Visdom with `use_incoming_socket=False`, which will prevent waiting for this request to timeout.
    
    
    Could not connect to Visdom server.
     Trying to start a server....
    Command: /home/xingxiangrui/env/bin/python -m visdom.server -p 8097 &>/dev/null &
    create web directory ./checkpoints/norText_2_cotton_cyclegan/web...

# 二、训练命令行 #

## 2.1 norText\_2\_cotton ##

`注意，在哪里运行，checkpoints就存在什么地方，导致checkpoints存在home/xingxiangrui目录下`

## 2.2 norText\_2\_falText ##

env/bin/python /home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/train.py --dataroot /home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/datasets/norText\_2\_falText --name norText\_2\_falText\_cyclegan --model cycle\_gan `--no_html`

`--continue_train`

## `2.3 norText_2_tear` ##

env/bin/python /home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/train.py --dataroot /home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/datasets/norText\_2\_tear --name nor2tear\_cyclegan --model cycle\_gan `--no_html`

# 三、测试 #

## 3.1 测试相关命令行及含义 ##

python test.py --dataroot ./datasets/maps --name maps_cyclegan --model cycle_gan

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.

env/bin/python /home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/test.py --dataroot /home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/datasets/norText\_2\_cotton --name norText\_2\_cotton\_cyclegan --model cycle\_gan

### 两点注意 ###

不要在norText\_2\_cotton后面加testA文件夹，不然模型会出错。(换用另一种命令行就没事，用--model test)

File "/home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/data/unaligned_dataset.py", line 29, in __init__
        self.A_paths = sorted(make_dataset(self.dir_A, opt.max_dataset_size))   # load images from '/path/to/data/trainA'
      File "/home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/data/image_folder.py", line 25, in make_dataset
        assert os.path.isdir(dir), '%s is not a valid directory' % dir
    AssertionError: /home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/datasets/norText_2_cotton/testA/testA is not a valid directory

也不要让testB文件夹为空

File "/home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/data/unaligned_dataset.py", line 53, in __getitem__
        index_B = index % self.B_size
    ZeroDivisionError: integer division or modulo by zero

## 3.2 测试 ##

### 模型普通到棉上（both sides） ###

但是，这个是both sides，不是one side，one side命令行如下：

env/bin/python /home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/test.py --dataroot /home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/datasets/norText\_2\_cotton/testA --name norText\_2\_cotton\_cyclegan --model test --no\_dropout

但是one side的命令行之中，需要G.ph就是生成器，但是目录之中，只有G\_A.ph 和G\_B.ph这两种，不知道什么原因，我们需要再仔细研究下。

FileNotFoundError: \[Errno 2\] No such file or directory: './checkpoints/nor2tear\_cyclegan/latest\_net\_G.pth'

### 模型普通到撕裂上（both sides） ###

env/bin/python /home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/test.py --dataroot /home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/datasets/norText\_2\_tear --name nor2tear\_cyclegan --model cycle\_gan

普通到撕裂上（one side）

此命令行即根据dataroot之中的testA迁移到testB

[xingxiangrui@yq01-gpu-yq-face-21-5 xingxiangrui]$ env/bin/python /home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/test.py --dataroot /home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/datasets/norText_2_cotton --name norText_2_cotton_cyclegan --model cycle_gan
    ----------------- Options ---------------
                 aspect_ratio: 1.0
                   batch_size: 1
              checkpoints_dir: ./checkpoints
                    crop_size: 256
                     dataroot: /home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/datasets/norText_2_cotton	[default: None]
                 dataset_mode: unaligned
                    direction: AtoB
              display_winsize: 256
                        epoch: latest
                         eval: False
                      gpu_ids: 0
                    init_gain: 0.02
                    init_type: normal
                     input_nc: 3
                      isTrain: False                         	[default: None]
                    load_iter: 0                             	[default: 0]
                    load_size: 256
             max_dataset_size: inf
                        model: cycle_gan                     	[default: test]
                   n_layers_D: 3
                         name: norText_2_cotton_cyclegan     	[default: experiment_name]
                          ndf: 64
                         netD: basic
                         netG: resnet_9blocks
                          ngf: 64
                   no_dropout: True
                      no_flip: False
                         norm: instance
                        ntest: inf
                     num_test: 50
                  num_threads: 4
                    output_nc: 3
                        phase: test
                   preprocess: resize_and_crop
                  results_dir: ./results/
               serial_batches: False
                       suffix:
                      verbose: False
    ----------------- End -------------------
    dataset [UnalignedDataset] was created
    initialize network with normal
    initialize network with normal
    model [CycleGANModel] was created
    loading the model from ./checkpoints/norText_2_cotton_cyclegan/latest_net_G_A.pth
    loading the model from ./checkpoints/norText_2_cotton_cyclegan/latest_net_G_B.pth
    ---------- Networks initialized -------------
    [Network G_A] Total number of parameters : 11.378 M
    [Network G_B] Total number of parameters : 11.378 M
    -----------------------------------------------
    processing (0000)-th image... ['/home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/datasets/norText_2_cotton/testA/001.png']
    processing (0005)-th image... ['/home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/datasets/norText_2_cotton/testA/006.png']
    processing (0010)-th image... ['/home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/datasets/norText_2_cotton/testA/011.png']
    processing (0015)-th image... ['/home/xingxiangrui/pytorch-CycleGAN-and-pix2pix/datasets/norText_2_cotton/testA/016.png']

# 四、存储结果位置 #

因为运行的时的目录的位置就是实际结果存储的位置。

## 4.1 测试结果位置 ##

有个results文件夹，其中包含着相应的运行结果。

需要注意，testA中有x个文件，则结果也有x个组，每组中包括6个，fakeA，fakeB，realA，realB，recA，recB

### html ###

再results中与images并列的地方，存在index.html文件里面，其中包含了每种结果的汇总，（但是因为我们程序有一些问题，这部分暂时没有显示。）

## 4.2 训练结果位置 ##

运行目录下的/checkpoints/里面

# 五、loss常见值 #

## 5.1 norText\_2\_falText\_cyclegan ##

### 开始 ###

(epoch: 1, iters: 100, time: 0.893, data: 0.164) D_A: 0.774 G_A: 0.517 cycle_A: 1.244 idt_A: 0.813 D_B: 0.513 G_B: 0.355 cycle_B: 1.781 idt_B: 0.588
    (epoch: 1, iters: 200, time: 0.861, data: 0.002) D_A: 0.583 G_A: 0.436 cycle_A: 1.244 idt_A: 0.876 D_B: 0.418 G_B: 0.458 cycle_B: 1.429 idt_B: 0.694
    (epoch: 1, iters: 300, time: 0.905, data: 0.002) D_A: 0.286 G_A: 0.321 cycle_A: 1.381 idt_A: 0.670 D_B: 0.484 G_B: 0.760 cycle_B: 2.111 idt_B: 0.610
    (epoch: 1, iters: 400, time: 1.155, data: 0.002) D_A: 0.261 G_A: 0.473 cycle_A: 1.048 idt_A: 0.942 D_B: 0.288 G_B: 0.223 cycle_B: 2.142 idt_B: 0.459
    End of epoch 1 / 200 	 Time Taken: 419 sec
    learning rate = 0.0002000
    (epoch: 2, iters: 21, time: 0.867, data: 0.002) D_A: 0.278 G_A: 0.338 cycle_A: 0.865 idt_A: 0.598 D_B: 0.355 G_B: 0.113 cycle_B: 1.312 idt_B: 0.405
    (epoch: 2, iters: 121, time: 0.909, data: 0.002) D_A: 0.281 G_A: 0.363 cycle_A: 1.717 idt_A: 0.407 D_B: 0.286 G_B: 0.288 cycle_B: 0.917 idt_B: 0.968
    (epoch: 2, iters: 221, time: 0.864, data: 0.002) D_A: 0.249 G_A: 0.388 cycle_A: 0.993 idt_A: 0.722 D_B: 0.304 G_B: 0.582 cycle_B: 1.431 idt_B: 0.412
    (epoch: 2, iters: 321, time: 1.205, data: 0.002) D_A: 0.136 G_A: 0.259 cycle_A: 1.251 idt_A: 0.797 D_B: 0.171 G_B: 0.510 cycle_B: 2.232 idt_B: 0.632
    (epoch: 2, iters: 421, time: 0.857, data: 0.002) D_A: 0.283 G_A: 0.738 cycle_A: 0.740 idt_A: 0.875 D_B: 0.233 G_B: 0.340 cycle_B: 1.831 idt_B: 0.365
    End of epoch 2 / 200 	 Time Taken: 417 sec
    learning rate = 0.0002000
    (epoch: 3, iters: 42, time: 0.864, data: 0.002) D_A: 0.212 G_A: 0.329 cycle_A: 0.900 idt_A: 1.102 D_B: 0.226 G_B: 0.656 cycle_B: 2.429 idt_B: 0.436
    (epoch: 3, iters: 142, time: 0.886, data: 0.002) D_A: 0.214 G_A: 0.551 cycle_A: 1.099 idt_A: 0.471 D_B: 0.299 G_B: 1.530 cycle_B: 1.425 idt_B: 0.510
    (epoch: 3, iters: 242, time: 1.224, data: 0.002) D_A: 0.367 G_A: 0.347 cycle_A: 1.350 idt_A: 0.432 D_B: 0.360 G_B: 0.109 cycle_B: 1.078 idt_B: 0.655
    (epoch: 3, iters: 342, time: 0.869, data: 0.002) D_A: 0.328 G_A: 0.099 cycle_A: 1.522 idt_A: 0.450 D_B: 0.156 G_B: 0.546 cycle_B: 0.984 idt_B: 0.747
    (epoch: 3, iters: 442, time: 0.860, data: 0.002) D_A: 0.395 G_A: 0.295 cycle_A: 2.179 idt_A: 0.766 D_B: 0.377 G_B: 0.247 cycle_B: 1.586 idt_B: 1.249
    End of epoch 3 / 200 	 Time Taken: 417 sec
    learning rate = 0.0002000

epoch=35左右

learning rate = 0.0002000
    (epoch: 33, iters: 72, time: 0.865, data: 0.002) D_A: 0.090 G_A: 0.417 cycle_A: 0.420 idt_A: 0.237 D_B: 0.113 G_B: 0.361 cycle_B: 0.559 idt_B: 0.178
    (epoch: 33, iters: 172, time: 0.896, data: 0.002) D_A: 0.094 G_A: 0.302 cycle_A: 0.734 idt_A: 0.629 D_B: 0.354 G_B: 1.186 cycle_B: 1.013 idt_B: 0.311
    (epoch: 33, iters: 272, time: 1.217, data: 0.002) D_A: 0.312 G_A: 0.501 cycle_A: 0.481 idt_A: 0.275 D_B: 0.163 G_B: 0.178 cycle_B: 0.595 idt_B: 0.202
    (epoch: 33, iters: 372, time: 0.915, data: 0.002) D_A: 0.302 G_A: 0.279 cycle_A: 0.475 idt_A: 0.204 D_B: 0.091 G_B: 0.481 cycle_B: 0.465 idt_B: 0.187
    (epoch: 33, iters: 472, time: 0.863, data: 0.002) D_A: 0.214 G_A: 0.156 cycle_A: 0.393 idt_A: 0.239 D_B: 0.086 G_B: 1.344 cycle_B: 0.620 idt_B: 0.164
    End of epoch 33 / 200 	 Time Taken: 416 sec
    learning rate = 0.0002000
    (epoch: 34, iters: 93, time: 0.911, data: 0.002) D_A: 0.075 G_A: 0.455 cycle_A: 0.537 idt_A: 0.263 D_B: 0.180 G_B: 0.647 cycle_B: 0.701 idt_B: 0.248
    (epoch: 34, iters: 193, time: 1.169, data: 0.002) D_A: 0.239 G_A: 0.330 cycle_A: 0.757 idt_A: 0.299 D_B: 0.075 G_B: 0.334 cycle_B: 0.792 idt_B: 0.358
    (epoch: 34, iters: 293, time: 0.866, data: 0.002) D_A: 0.011 G_A: 0.313 cycle_A: 0.688 idt_A: 0.271 D_B: 0.121 G_B: 0.867 cycle_B: 0.591 idt_B: 0.305
    (epoch: 34, iters: 393, time: 0.863, data: 0.002) D_A: 0.163 G_A: 0.341 cycle_A: 0.372 idt_A: 0.224 D_B: 0.047 G_B: 0.663 cycle_B: 0.571 idt_B: 0.153
    End of epoch 34 / 200 	 Time Taken: 416 sec
    learning rate = 0.0002000
    (epoch: 35, iters: 14, time: 0.859, data: 0.002) D_A: 0.182 G_A: 0.433 cycle_A: 0.432 idt_A: 0.198 D_B: 0.397 G_B: 0.024 cycle_B: 0.455 idt_B: 0.206
    (epoch: 35, iters: 114, time: 1.158, data: 0.002) D_A: 0.187 G_A: 0.350 cycle_A: 0.603 idt_A: 0.291 D_B: 0.136 G_B: 0.152 cycle_B: 0.891 idt_B: 0.229
    (epoch: 35, iters: 214, time: 0.858, data: 0.002) D_A: 0.122 G_A: 0.542 cycle_A: 0.401 idt_A: 0.246 D_B: 0.180 G_B: 0.656 cycle_B: 0.552 idt_B: 0.174
    (epoch: 35, iters: 314, time: 0.892, data: 0.002) D_A: 0.256 G_A: 0.285 cycle_A: 0.423 idt_A: 0.205 D_B: 0.161 G_B: 0.227 cycle_B: 0.450 idt_B: 0.197
    (epoch: 35, iters: 414, time: 0.862, data: 0.002) D_A: 0.092 G_A: 0.535 cycle_A: 0.616 idt_A: 0.208 D_B: 0.082 G_B: 0.625 cycle_B: 0.504 idt_B: 0.316
    saving the model at the end of epoch 35, iters 16765
    End of epoch 35 / 200 	 Time Taken: 417 sec
    learning rate = 0.0002000

# 六、命令行汇总 #

### lambda 40 ###

训练

测试

### lambda 20 idt =0 ###

训练

测试

## 四显卡机 ##

### load 496,crop 256,GPU 0 ###

###   训练 ###

python train.py --dataroot datasets/single2poly-OK-dataset --name single2poly\_OK\_cyclegan --model cycle\_gan --no\_html``\--max\_dataset\_size 1000 `--preprocess scale_width_and_crop --load_size 496 --crop_size 256`\--gpu\_ids 0

###   测试 ###

python test.py --dataroot datasets/single2poly-OK-dataset --name single2poly\_OK\_cyclegan --model cycle\_gan --num\_test 200 `--preprocess scale_width --load_size 496`\--gpu\_ids 1

## lambda\_5,idt 0,crop 440,gpu 0 ##

###   训练 ###

none先试一下 即496尺寸直接送入网络，显存溢出

python train.py --dataroot datasets/single2poly-OK-dataset --name single2poly\_OK\_cyclegan\_lambda5idt0 --model cycle\_gan --no\_html``\--max\_dataset\_size 200 `--preprocess none`\--gpu\_ids 0 --lambda\_A 5 --lambda\_B 5 --lambda\_identity 0

不进行resize，直接crop为256，可以运行，逐渐将crop的尺寸增大为400正常 440可以采用此尺寸  480显存溢出,训练一次大概12小时

python train.py --dataroot datasets/single2poly-OK-dataset --name single2poly\_OK\_cyclegan\_lambda5idt0 --model cycle\_gan --no\_html``\--max\_dataset\_size 200 `--preprocess crop --load_size 496 --crop_size 440`\--gpu\_ids 0 --lambda\_A 5 --lambda\_B 5 --lambda\_identity 0

测试

python test.py --dataroot datasets/single2poly-OK-dataset --name single2poly\_OK\_cyclegan\_lambda5idt0 --model cycle\_gan --num\_test 100 `--preprocess none`\--gpu\_ids 0

### lambda\_2, idt 0 ###

###   训练 ###

python train.py --dataroot datasets/single2poly-OK-dataset --name single2poly\_OK\_cyclegan\_lambda2idt0 --model cycle\_gan --no\_html``\--max\_dataset\_size 200 `--preprocess crop --load_size 496 --crop_size 440`\--gpu\_ids 1 --lambda\_A 2 --lambda\_B 2 --lambda\_identity 0

测试

python test.py --dataroot datasets/single2poly-OK-dataset --name single2poly\_OK\_cyclegan\_lambda2idt0 --model cycle\_gan --num\_test 100 `--preprocess none`\--gpu\_ids 1

### lambda0.5,idt0 ###

###   训练 ###

python train.py --dataroot datasets/single2poly-OK-dataset --name single2poly\_OK\_cyclegan\_lambda0.5idt0 --model cycle\_gan --no\_html``\--max\_dataset\_size 200 `--preprocess crop --load_size 496 --crop_size 440`\--gpu\_ids 2 --lambda\_A 0.5 --lambda\_B 0.5 --lambda\_identity 0

测试

python test.py --dataroot datasets/single2poly-OK-dataset --name single2poly\_OK\_cyclegan\_lambda0.5idt0 --model cycle\_gan --num\_test 100 `--preprocess none`\--gpu\_ids 2

### lambda0 ###

###   训练 ###

测试

python test.py --dataroot datasets/single2poly-OK-dataset --name single2poly\_OK\_cyclegan\_lambda0 --model cycle\_gan --num\_test 100 `--preprocess none`\--gpu\_ids 3

### 测试流程 ###

上传文件到服务器

cd ....

测试前清空上次运行数据

rm -rf /home/xingxiangrui/results/norText\_2\_cotton\_cyclegan/test\_latest/

另一个终端测试

第一个终端打包文件（进入到文件夹打包，不然只在根目录生成，麻烦）

zip -r nor\_2\_cott\_\*\*\_\*\*.zip /home/xingxiangrui/results/norText\_2\_cotton\_cyclegan/test\_latest/images

下载到电脑

sz /home/xingxiangrui/results/norText\_2\_cotton\_cyclegan/test\_latest/\*\*.zip

[https_github.com_junyanz_pytorch-CycleGAN-and-pix2pix_blob_master_docs_datasets.md]: https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix/blob/master/docs/datasets.md
[visdom]: https://github.com/facebookresearch/visdom