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
916 stars 169 forks source link

nice work, about clip_len #27

Open loongofqiao opened 3 years ago

loongofqiao commented 3 years ago

您好,我想问下,论文中提及的video clip,具体是多长的片段呢,在训练的时候,您是每次迭代读取batch_size个视频帧片段吗?因为我看到您代码里有select_frames函数,相关的参数是clip_len和generate_interval?我对这个没太理解,训练的时候这两个怎么设置参数呢?期待您的回复,谢谢

Hangz-nju-cuhk commented 3 years ago

比如clip_len设置成16就会在dataloader里读取16帧,但是generate_interval设置为4的时候,在生成的时候只会以step 4选取4帧,以减少显存占用。

hkchae96 commented 2 years ago

Could you tell exactly what values you used for hyper-parmeters --clip_len \ --generate_interval \ --batchSize \ for training script?