3DSfMFaceReconstruction icon indicating copy to clipboard operation
3DSfMFaceReconstruction copied to clipboard

Published 20 hours ago • verified publisher icon BoomStarcuc

image sequences

Open SFM2020 opened this issue 2 years ago • 3 comments

How to input multi-view images when infering? It looks like we can only input a single view image to get a disparity map in the inference.py

SFM2020 avatar May 16 '22 14:05 SFM2020

Yes, I have the same question. Even when a directory containing an image sequence is fed into the program, it treats each image as a separate one. It would be nice to update the landmark positions accordingly. I am not sure that averaging would be best to do: say, the nose form is best revealed in the side view, so more weight should be given to the landmarks on the boundary of the face. I could not yet find how to obtain the estimated head rotation information from fitting. Maybe the author could help us here?

baba-yaga avatar Dec 14 '22 12:12 baba-yaga

Did you slove this problem?

Leung-GD avatar Apr 13 '23 03:04 Leung-GD

Hi, I have not been working on it. I find this project abandoned since the author provides no feedback and the current code just not working for me.

sergei9838 avatar Apr 13 '23 06:04 sergei9838