Co-Fusion4D: Spatio-temporal Collaborative Fusion for Robust 3D Object Detection
The paper presents Co-Fusion4D, a new framework designed to enhance 3D object detection in autonomous driving. It addresses issues related to spatiotemporal inconsistencies and feature misalignment in existing BEV-based detectors. The proposed method achieves state-of-the-art performance on the nuScenes benchmark without relying on external data or test-time augmentation.
- ▪Co-Fusion4D preserves cross-frame spatiotemporal consistency and suppresses temporal feature drift.
- ▪The framework employs a current-frame-centric strategy, selectively incorporating historical frames after filtering and alignment.
- ▪Co-Fusion4D integrates a Dual Attention Fusion module to improve spatiotemporal feature interaction.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Computer Vision and Pattern Recognition arXiv:2605.20301 (cs) [Submitted on 19 May 2026] Title:Co-Fusion4D: Spatio-temporal Collaborative Fusion for Robust 3D Object Detection Authors:Wenxuan Li, Qin Zou, Shoubing Chen, Chi Chen, Yingyi Yang, Shoubing Chen, Qingxiang Meng View a PDF of the paper titled Co-Fusion4D: Spatio-temporal Collaborative Fusion for Robust 3D Object Detection, by Wenxuan Li and 6 other authors View PDF HTML (experimental) Abstract:In autonomous driving, 3D object detection is essential for accurate perception and reliable decision-making. However, object motion and ego-motion often induce cross-frame spatiotemporal inconsistencies in BEV-based detectors, leading to temporal BEV feature misalignment and degraded spatiotemporal consistency.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
