search

tijiang13 / InstantAvatar

InstantAvatar: Learning Avatars from Monocular Video in 60 Seconds (CVPR 2023)
365 stars 31 forks source link

Why I get ghosting for own dataset? #80

Open ZhenyanSun opened 2 weeks ago

ZhenyanSun commented 2 weeks ago

![Uploading 1.png…]()

tijiang13 commented 2 weeks ago

Hi Zhenyan,

Judging from the output, I assume the estimated SMPL is not very accurate for the legs. A few questions:

  1. did you run the preprocessing pipeline successfully? Normally by refining the SMPL with Openpose detection, the alignment will be good enough to get rid of this kind of noise.
  2. Alternatively, you can try to replace ROMP with 4DHuman using the pipeline I shared in the other thread https://github.com/tijiang13/InstantAvatar/issues/79

Best, Tianjian