Title:
Enhancing 3D Object Detection with Point Diffusion

Abstract:
Traditional 3D object detection models typically approach the problem by combining the tasks of object classification and shape regression into a unified end-to-end learning framework.
However, one-shot object detectors
In this paper, we propose a novel approach to refine 3D object detection results using point diffusion.
Instead of proposing
Extensive experiments on Lyft, NuScenes, and Ithaca-365 datasets demonstrate our method's efficacy in significantly enhancing object detection accuracy