search

cleardusk / 3DDFA_V2

The official PyTorch implementation of Towards Fast, Accurate and Stable 3D Dense Face Alignment, ECCV 2020.
MIT License
2.89k stars 511 forks source link

the detail of 3D Aided Short-video-synthesis? #107

Open wujinke opened 3 years ago

wujinke commented 3 years ago

有几个问题想请教下？ 1、短视频合成这个操作是在训练阶段加入的吗？ 2、论文中说batch-level的，是先将原始数据划分batch，然后在每一个batch中将每一个图片生成n帧视频帧？还是先根据每个图象生成视频帧，再划分batch训练。

prokaryote-hgy commented 1 year ago

@wujinke 我也有同样的问题，请问你解决这个疑问了吗

JH95-ai commented 1 year ago

I think it is the same as Repeated Augmentation