search

qiuyu96 / CoDeF

[CVPR'24 Highlight] Official PyTorch implementation of CoDeF: Content Deformation Fields for Temporally Consistent Video Processing
https://qiuyu96.github.io/CoDeF/
Other
4.84k stars 385 forks source link

Question about Canonical Step #52

Closed CharmsGraker closed 1 year ago

CharmsGraker commented 1 year ago

Thanks for great work! I notice all images in transformed video are warped by the first transformed img], aka canonical_img, which makes me little confused. code link I'm wondering the case like people turning around could be well handled by this field.

ken-ouyang commented 1 year ago

I think the turning around case might not work well with only one canonical presented. We are currently working on the improved version specifically designed for human. Please stay tuned.