vimalabs/VIMABench

Depth image

XihangYU630 opened this issue · 2 comments

It says that assert "depth" not in modalities, "FIXME: fix depth normalization". Can we change the modality to include "depth"?

After adding depth in modality and comment out "depth" not in modalities, "FIXME: fix depth normalization", there is another error saying assert self.observation_space.contains(obs)

Thanks for playing with our codebase!
Please note that our models do not require depth modality during training, therefore the codebase may not support it out of the box. IIRC, the issue is about the modified camera setting such that returned depth values are within a weird range. It may require some modifications to fix. Contributions from the community is welcomed!