【AI】CycleGan对抗生成网络遥感影像生成地图效果测试
今天看到一个有趣的项目，CycleGan对抗生成网络把马生成成斑马，还有一个测试用例是用遥感影像生成平面地图的效果，效果如下图所示，我大学是遥感专业，看到遥感影像就触动了我的原神，于是原神启动，肝一个测试的玩玩。
源码地址：https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix
0.准备工作
其实按照官方文档玩，没有什么特别需要准备的，只是为了防止遗忘，多赘述一下了。
项目有提供数据集和预训练模型，这种最适合懒人操作了。
-
下载数据集
其实源码中已经放了下载数据集的脚本，只需要运行脚本即可，但是有时候下载不稳定，我们还是提前下载下来放在环境下面最稳妥。
下载地址：http://efrosgans.eecs.berkeley.edu/cyclegan/datasets
解压后放在项目的dataset目录下面
-
下载预训练模型
下载地址：http://efrosgans.eecs.berkeley.edu/cyclegan/pretrained_models/
预训练模型下载之后需要改一下名字，在项目的checkpoint目录下面创建预训练模型的文件夹，并把下载的预训练模型改名为latest_net_G.pth
3.安装依赖
我这边使用的Anaconda虚拟环境进行运行，选择一个之前安装了pytorch框架的虚拟环境，然后再终端中切换到项目文件夹下面，使用pip install -r requirments.txt安装项目依赖。
项目依赖挺简单，除了pytorch的包外，最主要的就是两个包了，因为我的虚拟环境上已经安装了pytorch的依赖，所以前两个依赖可以注释掉
#torch>=0.4.1
#torchvision>=0.2.1
dominate>=2.3.1
visdom>=0.1.8.3
1.运行测试
安装完就可以运行了，直接在项目的根目录下面运行即可
python test.py --dataroot datasets/maps/testA --name sat2map_pretrained --model test --no_dropout
参数介绍：
--dataroot 测试数据集存放的路径
--name 模型名称，这里跟checkpoint目录下面的文件夹名称一致
--model test 使用test参数说明我们是在测试
终端输出内容展示：
(pytorch) PS F:\AIStudy\pytorch-CycleGAN-and-pix2pix> python test.py --dataroot datasets/maps/testA --name sat2map_pretrained --model test --no_dropout
----------------- Options ---------------
aspect_ratio: 1.0
batch_size: 1
checkpoints_dir: ./checkpoints
crop_size: 256
dataroot: datasets/maps/testA [default: None]
dataset_mode: single
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: test
model_suffix:
n_layers_D: 3
name: sat2map_pretrained [default: experiment_name]
ndf: 64
netD: basic
netG: resnet_9blocks
ngf: 64
no_dropout: True [default: False]
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 [SingleDataset] was created
[ReflectionPad2d((3, 3, 3, 3)), Conv2d(3, 64, kernel_size=(7, 7), stride=(1, 1)), InstanceNorm2d(64, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False), ReLU(inplace=True), Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1)), InstanceNorm2d(128, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False), ReLU(inplace=True), Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1)), InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False), ReLU(inplace=True), ResnetBlock(
(conv_block): Sequential(
(0): ReflectionPad2d((1, 1, 1, 1))
(1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(2): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
(3): ReLU(inplace=True)
(4): ReflectionPad2d((1, 1, 1, 1))
(5): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(6): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
)
), ResnetBlock(
(conv_block): Sequential(
(0): ReflectionPad2d((1, 1, 1, 1))
(1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(2): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
(3): ReLU(inplace=True)
(4): ReflectionPad2d((1, 1, 1, 1))
(5): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(6): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
)
), ResnetBlock(
(conv_block): Sequential(
(0): ReflectionPad2d((1, 1, 1, 1))
(1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(2): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
(3): ReLU(inplace=True)
(4): ReflectionPad2d((1, 1, 1, 1))
(5): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(6): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
)
), ResnetBlock(
(conv_block): Sequential(
(0): ReflectionPad2d((1, 1, 1, 1))
(1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(2): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
(3): ReLU(inplace=True)
(4): ReflectionPad2d((1, 1, 1, 1))
(5): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(6): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
)
), ResnetBlock(
(conv_block): Sequential(
(0): ReflectionPad2d((1, 1, 1, 1))
(1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(2): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
(3): ReLU(inplace=True)
(4): ReflectionPad2d((1, 1, 1, 1))
(5): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(6): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
)
), ResnetBlock(
(conv_block): Sequential(
(0): ReflectionPad2d((1, 1, 1, 1))
(1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(2): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
(3): ReLU(inplace=True)
(4): ReflectionPad2d((1, 1, 1, 1))
(5): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(6): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
)
), ResnetBlock(
(conv_block): Sequential(
(0): ReflectionPad2d((1, 1, 1, 1))
(1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(2): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
(3): ReLU(inplace=True)
(4): ReflectionPad2d((1, 1, 1, 1))
(5): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(6): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
)
), ResnetBlock(
(conv_block): Sequential(
(0): ReflectionPad2d((1, 1, 1, 1))
(1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(2): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
(3): ReLU(inplace=True)
(4): ReflectionPad2d((1, 1, 1, 1))
(5): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(6): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
)
), ResnetBlock(
(conv_block): Sequential(
(0): ReflectionPad2d((1, 1, 1, 1))
(1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(2): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
(3): ReLU(inplace=True)
(4): ReflectionPad2d((1, 1, 1, 1))
(5): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(6): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
)
)]
[ReflectionPad2d((3, 3, 3, 3)), Conv2d(3, 64, kernel_size=(7, 7), stride=(1, 1)), InstanceNorm2d(64, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False), ReLU(inplace=True), Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1)), InstanceNorm2d(128, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False), ReLU(inplace=True), Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1)), InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False), ReLU(inplace=True), ResnetBlock(
(conv_block): Sequential(
(0): ReflectionPad2d((1, 1, 1, 1))
(1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(2): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
(3): ReLU(inplace=True)
(4): ReflectionPad2d((1, 1, 1, 1))
(5): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(6): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
)
), ResnetBlock(
(conv_block): Sequential(
(0): ReflectionPad2d((1, 1, 1, 1))
(1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(2): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
(3): ReLU(inplace=True)
(4): ReflectionPad2d((1, 1, 1, 1))
(5): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(6): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
)
), ResnetBlock(
(conv_block): Sequential(
(0): ReflectionPad2d((1, 1, 1, 1))
(1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(2): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
(3): ReLU(inplace=True)
(4): ReflectionPad2d((1, 1, 1, 1))
(5): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(6): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
)
), ResnetBlock(
(conv_block): Sequential(
(0): ReflectionPad2d((1, 1, 1, 1))
(1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(2): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
(3): ReLU(inplace=True)
(4): ReflectionPad2d((1, 1, 1, 1))
(5): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(6): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
)
), ResnetBlock(
(conv_block): Sequential(
(0): ReflectionPad2d((1, 1, 1, 1))
(1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(2): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
(3): ReLU(inplace=True)
(4): ReflectionPad2d((1, 1, 1, 1))
(5): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(6): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
)
), ResnetBlock(
(conv_block): Sequential(
(0): ReflectionPad2d((1, 1, 1, 1))
(1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(2): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
(3): ReLU(inplace=True)
(4): ReflectionPad2d((1, 1, 1, 1))
(5): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(6): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
)
), ResnetBlock(
(conv_block): Sequential(
(0): ReflectionPad2d((1, 1, 1, 1))
(1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(2): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
(3): ReLU(inplace=True)
(4): ReflectionPad2d((1, 1, 1, 1))
(5): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(6): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
)
), ResnetBlock(
(conv_block): Sequential(
(0): ReflectionPad2d((1, 1, 1, 1))
(1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(2): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
(3): ReLU(inplace=True)
(4): ReflectionPad2d((1, 1, 1, 1))
(5): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(6): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
)
), ResnetBlock(
(conv_block): Sequential(
(0): ReflectionPad2d((1, 1, 1, 1))
(1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(2): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
(3): ReLU(inplace=True)
(4): ReflectionPad2d((1, 1, 1, 1))
(5): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))
(6): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
)
), ConvTranspose2d(256, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), output_padding=(1, 1)), InstanceNorm2d(128, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False), ReLU(inplace=True), ConvTranspose2d(128, 64, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), output_padding=(1, 1)), InstanceNorm2d(64, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False), ReLU(inplace=True), ReflectionPad2d((3, 3, 3, 3)), Conv2d(64, 3, kernel_size=(7, 7), stride=(1, 1)), Tanh()]
initialize network with normal
model [TestModel] was created
loading the model from ./checkpoints\sat2map_pretrained\latest_net_G.pth
---------- Networks initialized -------------
[Network G] Total number of parameters : 11.378 M
-----------------------------------------------
creating web directory ./results/sat2map_pretrained\test_latest
processing (0000)-th image... ['datasets/maps/testA\\1000_A.jpg']
processing (0005)-th image... ['datasets/maps/testA\\1005_A.jpg']
processing (0010)-th image... ['datasets/maps/testA\\100_A.jpg']
processing (0015)-th image... ['datasets/maps/testA\\1014_A.jpg']
processing (0020)-th image... ['datasets/maps/testA\\1019_A.jpg']
processing (0025)-th image... ['datasets/maps/testA\\1023_A.jpg']
processing (0030)-th image... ['datasets/maps/testA\\1028_A.jpg']
processing (0035)-th image... ['datasets/maps/testA\\1032_A.jpg']
processing (0040)-th image... ['datasets/maps/testA\\1037_A.jpg']
processing (0045)-th image... ['datasets/maps/testA\\1041_A.jpg']
这里打印了模型参数、模型结构以及处理进度。
2.结果展示
执行结束后会在模型的result路径下（results\sat2map_pretrained\test_latest）生成查看样例，点击打开index.html就可以查看模型处理的效果
效果如下：
是不是很有趣！