Skip to content

Point Cloud Covers 360° but Semantic Labels Only Take Front Half, Mismatch Between Feature and Label Space? #93

@zhangcj13

Description

@zhangcj13

Thank you very much for your code.
I would like to ask a question regarding BEV semantic segmentation. The point cloud used for feature extraction covers the front, back, left, and right directions, but the segmentation labels only take the front half (x in 0 to 32m).
During training, the extracted features are not truncated before computing the loss. I'm wondering if this would cause the feature extraction to focus only on the front part of the information? Is this intentionally designed ?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions