Topology- and Gradient-Guided Knowledge Distillation for Point Cloud Semantic Segmentation

TMLR Paper7224 Authors

Abstract: Point cloud processing has gained significant attention due to its critical role in applications such as autonomous driving and 3D object recognition. However, deploying high-performance models like Point Transformer V3 in resource-constrained environments remains challenging due to their high computational and memory demands. This work introduces a novel distillation framework that leverages topology-aware representations and gradient-guided knowledge distillation to effectively transfer knowledge from a high-capacity teacher to a lightweight student model. Our approach captures the underlying geometric structures of point clouds while selectively guiding the student model's learning process through gradient-based feature alignment. Experimental results in the Nuscenes, SemanticKITTI, and Waymo datasets demonstrate that the proposed method achieves competitive performance, with an approximately 16$\times$ reduction in model size and up to 1.9$\times$ decrease in inference time compared to its teacher model. Notably, on NuScenes, our method achieves competitive performance among knowledge distillation techniques trained solely on LiDAR data, surpassing prior knowledge distillation baselines in segmentation performance. Our implementation is available anonymously at https://anonymous.4open.science/r/PTv3-distill-4E9E

Submission Type: Regular submission (no more than 12 pages of main content)

Changes Since Last Submission: N/A

Assigned Action Editor: ~Adam_W_Harley1

Submission Number: 7224