Skip to content

Issue with point cloud loss during processing in transform config #41

Description

@sumuru789

I have encountered an issue when using your project, where some points in the point cloud are being lost during processing. Specifically, in the config file, the transform configuration is set as follows:transform=[ dict(type="NormalizeCoord"), dict(type="CenterShift", apply_z=True), dict( type="GridSample", grid_size=0.01, keys=("coord", "color", "normal", "origin_coord", "face_index"), hash_type="fnv", mode="train", return_grid_coord=True, return_inverse=True, ), dict(type="CenterShift", apply_z=False), dict(type="NormalizeColor"), dict(type="ToTensor"), dict( type="Collect", keys=("coord", "grid_coord", "inverse", "origin_coord", "face_index"), feat_keys=("coord", "normal", "color"), ), ] It seems that during the point cloud processing, some points are discarded, especially when the GridSample operation is applied. Could you clarify which part of this configuration might be causing the points to be lost? How can I adjust the transform settings to prevent the points from being discarded?

Also, if we decide to avoid discarding points, would it cause any issues with extracting features from dinov2?

Thanks in advance for your help!

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions