CaRaFFusion: Improving 2D Semantic Segmentation with Camera-Radar Point Cloud Fusion and Zero-Shot Image Inpainting

TL;DR

This paper introduces CaRaFFusion, a novel camera-radar fusion framework that enhances 2D semantic segmentation by integrating radar data, pseudo-masks, and inpainting techniques, significantly improving performance under adverse weather conditions.

Contribution

It presents a new fusion architecture that combines radar point features with diffusion-based inpainting to improve semantic segmentation robustness and accuracy.

Findings

Improves camera-only segmentation by 2.63% in mIoU.

Enhances fusion architecture by 1.48% in mIoU.

Effective under adverse weather conditions.

Abstract

Segmenting objects in an environment is a crucial task for autonomous driving and robotics, as it enables a better understanding of the surroundings of each agent. Although camera sensors provide rich visual details, they are vulnerable to adverse weather conditions. In contrast, radar sensors remain robust under such conditions, but often produce sparse and noisy data. Therefore, a promising approach is to fuse information from both sensors. In this work, we propose a novel framework to enhance camera-only baselines by integrating a diffusion model into a camera-radar fusion architecture. We leverage radar point features to create pseudo-masks using the Segment-Anything model, treating the projected radar points as point prompts. Additionally, we propose a noise reduction unit to denoise these pseudo-masks, which are further used to generate inpainted images that complete the missing information in the original images. Our method improves the camera-only segmentation baseline by 2.63% in mIoU and enhances our camera-radar fusion architecture by 1.48% in mIoU on the Waterscenes dataset. This demonstrates the effectiveness of our approach for semantic segmentation using camera-radar fusion under adverse weather conditions.