Hangz-nju-cuhk / Talking-Face_PC-AVS

Code for Pose-Controllable Talking Face Generation by Implicitly Modularized Audio-Visual Representation (CVPR 2021)
Creative Commons Attribution 4.0 International
919 stars 169 forks source link

关于测试集的测试结果 #17

Open wlj961012 opened 3 years ago

wlj961012 commented 3 years ago

目前,random选取了测试集的部分数据, 总体来说,就是生成的video<输入图片的清晰度,有些视频生成的比较清晰(例如倒数第二个),大部分有一些糊。现在,我不太确定这是否是正常的测试结果(目前,没有改动GitHub给出的inference里面的任何参数)。期待您的回复,谢谢。

https://user-images.githubusercontent.com/31840936/118944466-2c4e8900-b987-11eb-876e-74fb68862278.mp4

https://user-images.githubusercontent.com/31840936/118944684-5dc75480-b987-11eb-83c6-8a10c1727639.mp4

https://user-images.githubusercontent.com/31840936/118944919-95ce9780-b987-11eb-968f-5335bc53ebdc.mp4

https://user-images.githubusercontent.com/31840936/118945012-a717a400-b987-11eb-9ea0-ab5c80db4548.mp4

https://user-images.githubusercontent.com/31840936/118945375-fa89f200-b987-11eb-8755-76183c17c83f.mp4

Hangz-nju-cuhk commented 3 years ago

Hi, 这些应该属于高难度的case,结果不理想也是正常,还不知道有没有任何face reenact 或者face rotation的方式可以handle这些case。