facebookresearch / consistent_depth

We estimate dense, flicker-free, geometrically consistent depth from monocular video, for example hand-held cell phone video.
MIT License
1.61k stars 236 forks source link

Masking out dynamic object is not working. #43

Open Minisal opened 3 years ago

Minisal commented 3 years ago

The provided option about masking out dynamic object for camera pose estimation is not working.

main.py: error: unrecognized arguments: --mask_path

It seems that the project does not provide an option about mask. And the mask in the final results is based on two frames. The mask required in the README is based on a single frame.

Are there other ways to mask out dynamic objects?