DiffuBox: Refining 3D Object Detection with Point Diffusion

TL;DR

DiffuBox introduces a diffusion-based method for refining 3D object detection boxes, improving localization accuracy across different domains and sensor setups by leveraging a domain-agnostic model conditioned on LiDAR data.

Contribution

The paper presents a novel diffusion-based refinement approach that enhances 3D object detection robustness under domain shifts, outperforming existing methods in various settings.

Findings

Significant improvements in localization accuracy across datasets.

Effective domain adaptation with diffusion-based refinement.

Applicable to multiple object classes and detection models.

Abstract

Ensuring robust 3D object detection and localization is crucial for many applications in robotics and autonomous driving. Recent models, however, face difficulties in maintaining high performance when applied to domains with differing sensor setups or geographic locations, often resulting in poor localization accuracy due to domain shift. To overcome this challenge, we introduce a novel diffusion-based box refinement approach. This method employs a domain-agnostic diffusion model, conditioned on the LiDAR points surrounding a coarse bounding box, to simultaneously refine the box's location, size, and orientation. We evaluate this approach under various domain adaptation settings, and our results reveal significant improvements across different datasets, object classes and detectors. Our PyTorch implementation is available at \href{https://github.com/cxy1997/DiffuBox}{https://github.com/cxy1997/DiffuBox}.