New issue

Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.

#

By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? # to your account

Jump to bottom

我尝试实现视频到视频的推理，结果很差 #79

Open

zhanghongyong123456 opened this issue Aug 24, 2024 · 1 comment

zhanghongyong123456 commented Aug 24, 2024 •

edited

Loading

1.我实现视频到视频的推理，将处理参考图像的代码，修改为处理视频帧代码，测试结果很差，有什么方法能提高效果吗？
这是视频到视频推理：

09.mp4

这是图片到视频的推理效果：

10.mp4

这是我代码修改主要地方，请大佬指点

对于函数 prepare_batch_from_inp
遍历参考视频每一帧合并得到sample字典信息：
修改函数 get_driving_motion ，参照drv_secc_colors 处理方法，处理了 src_secc_colors和cano_secc_colors ，

lvZic commented Dec 12, 2024

hi, have u figured it out?

# for free to join this conversation on GitHub. Already have an account? # to comment