facebookresearch / facebookresearch/dynamic_stereo
Completely Broken Depth
Open
- Dominant language
- Jupyter Notebook
- Stars
- 244
- Forks
- 17
- PR merge metrics
- No merged PRs in 30d
Description
Hey there,
Sometimes, the depth and disparity map are completely broken. However, the `valid_disp` mask still has non-zero values.
For example, 0 frame of`56aada-7_obj` depth map is incorrect.
What is the most reliable way of filtering out such samples?
Contributor guide
Research direction
No files, tests, or entry points are identified in the issue. Start by reproducing the broken depth map for frame 0 of `56aada-7_obj` and compare it with the non-zero `valid_disp` mask. Done would require an agreed, reliable way to identify and filter such samples.
Written by the indexing model from the issue text.
Assessment
- Domain
- computer-vision
- Issue type
- Bug
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100