Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

parse_visual_obs_mode_to_struct doesn't support obs_mode="sensor_data" #605

Closed
hesic73 opened this issue Oct 5, 2024 · 1 comment · Fixed by #606
Closed

parse_visual_obs_mode_to_struct doesn't support obs_mode="sensor_data" #605

hesic73 opened this issue Oct 5, 2024 · 1 comment · Fixed by #606

Comments

@hesic73
Copy link
Contributor

hesic73 commented Oct 5, 2024 •
edited
Loading

ManiSkill/mani_skill/envs/sapien_env.py

Line 114 in 362d43d

SUPPORTED_OBS_MODES = ("state", "state_dict", "none", "sensor_data", "rgb", "depth", "segmentation", "rgbd", "rgb+depth", "rgb+depth+segmentation", "rgb+segmentation", "depth+segmentation", "pointcloud")

ManiSkill/mani_skill/envs/utils/observations/__init__.py

Lines 14 to 53 in 362d43d

def parse_visual_obs_mode_to_struct(obs_mode: str) -> CameraObsTextures:
"""Given user supplied observation mode, return a struct with the relevant textures that are to be captured"""
if obs_mode == "rgb":
return CameraObsTextures(
rgb=True, depth=False, segmentation=False, position=False
)
elif obs_mode == "rgbd":
return CameraObsTextures(
rgb=True, depth=True, segmentation=False, position=False
)
elif obs_mode == "depth":
return CameraObsTextures(
rgb=False, depth=True, segmentation=False, position=False
)
elif obs_mode == "segmentation":
return CameraObsTextures(
rgb=False, depth=False, segmentation=True, position=False
)
elif obs_mode == "rgb+depth":
return CameraObsTextures(
rgb=True, depth=True, segmentation=False, position=False
)
elif obs_mode == "rgb+depth+segmentation":
return CameraObsTextures(
rgb=True, depth=True, segmentation=True, position=False
)
elif obs_mode == "rgb+segmentation":
return CameraObsTextures(
rgb=True, depth=False, segmentation=True, position=False
)
elif obs_mode == "depth+segmentation":
return CameraObsTextures(
rgb=False, depth=True, segmentation=True, position=False
)
elif obs_mode == "pointcloud":
return CameraObsTextures(
rgb=True, depth=False, segmentation=True, position=True
)
else:
return None

The 3D Diffusion Policy seems to require not only point cloud data and RGB images but also depth information, which obs_mode="pointcloud" doesn't support. Therefore, I expect the return value for obs_mode="sensor_data" should be:

CameraObsTextures(
            rgb=True, depth=True, segmentation=True, position=True
        )

By the way, in the point cloud representation xyzw, what does the w stand for?
@StoneT2000
Copy link
Member

StoneT2000 commented Oct 5, 2024 •
edited
Loading

w stand for infinite points, it's a mask. 0 is infinity points and shouldn't be used if it's a pointcloud / depth datapoint

Your fix is correct, feel free to make a PR to fix (currently away so can't get to it now)

@hesic73 hesic73 mentioned this issue Oct 5, 2024
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging a pull request may close this issue.

add missing 'sensor_data' handling in parse_visual_obs_mode_to_struct hesic73/ManiSkill
2 participants
@StoneT2000 @hesic73