Missing depth and semantic modalities

#1
by usermech1 - opened

Hi Staer Robotics Team,
I am reaching out because there seems to be a discrepancy between the documentation and the actual files currently hosted on the server under the v0.2 tag. The dataset landing page highlight the inclusion of 16-bit inverse depth maps, semantic segmentation, and instance segmentation maps.
However, looking at the repository file tree for scene01, the walkthrough directories (e.g., scene01/walkthroughs/001/cam0/) only contain:

  • poses.csv
  • sensor.yaml
  • video.mp4
  • preview.mp4
    The raw, uncompressed folders for depth0/ or semantic0/ do not appear to be present on the Hugging Face server. Is this an omission in the current v0.2 rolling release upload, or are the dense per-pixel depth and semantic maps isolated under an unlisted configuration/branch that requires specific download flags?

Dear Staer Robotics Team:
I am also noticing this issue. Is it a thing of Hugging Face or of the uploaded data? I am interested in scene 6 but in the preview here i cannot see more than cam_0 and the images are sadly not there. Any answer to this thread?

Sign up or log in to comment