first-order-model-master（avatarify一阶动态模型）代码解说（蚂蚁呀嘿 附完整运行代码）

1.按网上avatarify代码，first-order-model-maste解读，但是缺少运行的必要程序(vox-adv-cpk.pth.tar)。

（参考(https://blog.csdn.net/csdnnews/article/details/108570938)可以解决文件缺失问题）

运行环境：Python 3.7   

C:\Users\Administrator\Desktop> python
Python 3.7.2 (tags/v3.7.2:9a3ffc0492, Dec 23 2018, 23:09:28) [MSC v.1916 64 bit (AMD64)] on win32

需要的库文件如下：

如果报错误，就pip install XXX 自行安装。

C:\Users\Administrator>pip freeze
certifi==2020.12.5
chardet==4.0.0
cycler==0.10.0
decorator==4.4.2
face-alignment==1.3.3
idna==2.10
imageio==2.9.0
imageio-ffmpeg==0.4.3
joblib==1.0.1
kiwisolver==1.3.1
llvmlite==0.35.0
matplotlib==3.3.4
msgpack==1.0.2
msgpack-numpy==0.4.7.1
networkx==2.5
numba==0.52.0
numpy==1.20.1
opencv-python==4.2.0.34
pandas==1.2.3
Pillow==8.1.1
pyfakewebcam==0.1.0
pyparsing==2.4.7
python-dateutil==2.8.1
pytz==2021.1
PyWavelets==1.1.1
PyYAML==5.3.1
pyzmq==20.0.0
requests==2.25.1
scikit-image==0.18.1
scikit-learn==0.24.1
scipy==1.6.1
six==1.15.0
sklearn==0.0
threadpoolctl==2.1.0
tifffile==2021.3.4
torch==1.8.0
torchvision==0.9.0
tqdm==4.58.0
typing-extensions==3.7.4.3
urllib3==1.26.3

=================================================

下载速度慢，可以更改镜像点。做法如下：

在C:\Users\Administrator\pip目录下，修改pip.ini文件（直接copy文字即可）

[global]
index-url = http://pypi.douban.com/simple/
[install]
trusted-host = pypi.douban.com

======================================

2.又在网上搜到相关代码，运行如下。

(https://blog.csdn.net/csdnnews/article/details/108570938)

"怎样用 Python 控制图片人物动起来？一文就能 Get！"

   

3.GPU问题解决。

可参考如下 https://blog.csdn.net/m0_37690102/article/details/108364458

“解决问题：AssertionError: Torch not compiled with CUDA enabled”

此错误是由于下载的torch没有cuda,在运行时就会出错，经过查阅，在程序最开始的地方加上：
device = torch.device(“cuda” if torch.cuda.is_available() else “cpu”)
代码其余地方出现.cuda()的地方改成.to(device)就可以在无gpu的环境中运行了。

4,运行demo.py  

python demo.py

--config config/vox-adv-256.yaml

--driving_video path/to/driving/1.mp4

--source_image path/to/source/7.jpg

--checkpoint path/to/checkpoint/vox-adv-cpk.pth.tar

--relative

--adapt_scale

5.运行如下。

================================================================================================================================

C:\Python\Python37\lib\site-packages\torch\nn\functional.py:3500: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and now uses scale_factor directly, instead of relying on the computed output size. If you wish to restore the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details.
  "The default behavior for interpolate/upsample with float scale_factor changed "
  0%|                                                                                             | 0/211 [00:00<?, ?it/s]C:\Python\Python37\lib\site-packages\torch\nn\functional.py:3826: UserWarning: Default grid_sample and affine_grid behavior has changed to align_corners=False since 1.3.0. Please specify align_corners=True if the old behavior is desired. See the documentation of grid_sample for details.
  "Default grid_sample and affine_grid behavior has changed "
C:\Python\Python37\lib\site-packages\torch\nn\functional.py:3455: UserWarning: Default upsampling behavior when mode=bilinear is changed to align_corners=False since 0.4.0. Please specify align_corners=True if the old behavior is desired. See the documentation of nn.Upsample for details.
  "See the documentation of nn.Upsample for details.".format(mode)
C:\Python\Python37\lib\site-packages\torch\nn\functional.py:1709: UserWarning: nn.functional.sigmoid is deprecated. Use torch.sigmoid instead.
  warnings.warn("nn.functional.sigmoid is deprecated. Use torch.sigmoid instead.")
  0%|▍                  | 1/211 [00:38<2:13:00, 38.00s/it]  1%|▊         | 2/211 [01:12<2:04:46, 35.82s/it]  

1%|█▏                 | 3/211 [01:50<2:08:07, 36.96s/it  

2%|█▌              | 4/211 [02:24<2:03:57, 35.93s/it  2%|█▉       | 5/211 [02:59<2:01:57, 35.52s/it

 3%|██▎         | 6/211 =[03:33<1:59:49, 35.07s/i 3%|██▊     | 7/211 [04:08<1:58:12, 34.77s/i  

4%|███▏      | 8/211 [04:42<1:57:26, 34.71s/  4%|███▌    | 9/211 [05:16<1:56:04, 34.48s/  

5%|███▉     | 10/211 [05:50<1:55:15, 34.41s/  5%|████▎     | 11/211 [06:25<1:54:30, 34.35s

 6%|████▋       | 12/211 [06:59<1:53:50, 34.32s  6%|█████      | 13/211 [07:33<1:53:17, 34.33s  

7%|█████▍        | 14/211 [08:07<1:52:26, 34.24  7%|█████▊      | 15/211 [08:42<1:52:30, 34.44  

8%|██████▏   | 16/211 [09:18<1:53:11, 34.8  8%|██████▌       | 17/211 [09:54<1:53:24, 35.0  

9%|██████▉    | 18/211 [10:28<1:51:51, 34.7  9%|███████▍      | 19/211 [11:19<2:07:41, 39.  

9%|███████▊    | 20/211 [12:01<2:08:26, 40. 10%|████████▏  | 21/211 [12:35<2:02:17, 38

10%|████████▌     | 22/211 [13:09<1:57:18, 37 11%|████████▉    | 23/211 [13:44<1:53:57, 36.37s/it]

================================================================================================

6.结果

        

7.训练模型由文中的 茅佳源 源提供。

主页  https://jiayuanm.com/

7.有任何问题可留言，仅供学习参考。

==================================================================

代码下载地址：

链接：https://pan.baidu.com/s/1dlkJpYzT7tc-X3YgXTRtNA 
提取码：MNSY 

===================================================================

代码经修改，可以完整运行！

特别感谢

1.李秋键

(https://blog.csdn.net/csdnnews/article/details/108570938)

"怎样用 Python 控制图片人物动起来？一文就能 Get！"

提供的讲解思路，及其运行代码

2.Avatarify的开源代码提供者 Ali Aliev

Avatarify 的开源地址： https://github.com/alievk/avatarify

================================================================

后记：

效果达不到哈，本人也是在朋友用APP软件的恶搞下，开始研读源码。