You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi, I think the main reason is the z-axis ambiguity of the monocular pose estimation. The optimization process will further enlarge this error (Some other works may also meet the same issue, like https://huangyangyi.github.io/TeCH/,
). Moreover, the results of multi-view datasets evidence this hypothesis.
To solve this, I think a possible way you can try is to adjust the camera poses for training.
Hello @wyysf-98, Thank you for the work.
I was wondering why the Freeview rendering of people snapshot is slightly inclined. Is it really the issue with the monocular axis?
when I tried multi-view datasets, it works fine. but for monocular, it goes slightly inclined.
Would be great if you can mention anything about it?
The text was updated successfully, but these errors were encountered: