-
![eyes](https://user-images.githubusercontent.com/10509740/142052862-d0996feb-e927-4391-956c-641545802836.PNG)
Using wav2lip
No matter my source video, the eyes are blocky / pixelated / over sha…
ill13 updated
4 months ago
-
想训练压缩版的FOM模型,根据doc上的教程,先训练了kp_detector,最终收敛结果如下:
![image](https://user-images.githubusercontent.com/15072820/142101444-7471035b-fbbc-47b1-9dbf-47159917a4ee.png)
可以看到同比损失在0.2左右
然后继续训练generator,最终收敛结…
-
您好,我遇到两个问题,麻烦您看一下。
1. 这个模型下载不了![OMN}ZMGX8{CT(_4HFH998@I](https://user-images.githubusercontent.com/43515926/143401795-7e25df1e-aedf-48b9-b093-c082570e047d.png)
2. 在测试的时候模型加载的参数没看到在哪,请问这个模型路径应该放在哪啊?
…
-
我在跑paddlespeech里的一个[metaverse demo](https://github.com/PaddlePaddle/PaddleSpeech/blob/develop/demos/metaverse/run.sh),其中有用到paddlegan。运行过程中,paddlegan产生了报错:
```
~/PaddleSpeech/demos/metaverse$ ./run.s…
wwdok updated
4 months ago
-
看到运行示例中只有一个seed随机种子可以生成不同风格的图片,但是如果想要在一张基础图上增加一些其他的属性风格改变,需要怎么去做
-
I've used the example provided in the docs to clean and deoldify some old lectures using the following command:
`python tools/video-enhance.py --input you_video_path.mp4 --process_order DAIN DeOldify…
-
Which algorithms are the same as cyclegan, and the training data does not require unpaired pictures for image translation。
-
thanks for your contribution, i have sevel questions about starganv2.
1. If the length and width of the data in my training set are different, can it be modified in the program?
2. In addition to th…
-
-
I trained a model with [images256x256.tar](https://paddlegan.bj.bcebos.com/datasets/images256x256.tar).The ckpt.keys() are
`['gen', 'disc', 'gen_ema']`,
when stylegan2-ffhq-config-f.pd params.keys(…