Multi-View Attentive Contextualization for Multi-View 3D Object Detection

1 North Carolina State University
2 University of Central Florida
3 Ant Group 4 OPPO Seattle Research Center, USA
CVPR 2024

Abstract

We present Multi-View Attentive Contextualization (MvACon), a simple yet effective method for improving 2D-to-3D feature lifting in query-based multi-view 3D (MV3D) object detection. Despite remarkable progress witnessed in the field of query-based MV3D object detection, prior art often suffers from either the lack of exploiting high-resolution 2D features in dense attention-based lifting, due to high computational costs, or from insufficiently dense grounding of 3D queries to multi-scale 2D features in sparse attention-based lifting. Our proposed MvACon hits the two birds with one stone using a representationally dense yet computationally sparse attentive feature contextualization scheme that is agnostic to specific 2D-to-3D feature lifting approaches. In experiments, the proposed MvACon is thoroughly tested on the nuScenes benchmark, using both the BEVFormer and its recent 3D deformable attention (DFA3D) variant, as well as the PETR, showing consistent detection performance improvement, especially in enhancing performance in location, orientation, and velocity prediction. It is also tested on the Waymo-mini benchmark using BEVFormer with similar improvement. We qualitatively and quantitatively show that global cluster-based contexts effectively encode dense scene-level contexts for MV3D object detection. The promising results of our proposed MvACon reinforces the adage in computer vision --- ``(contextualized) feature matters".


MvACon


Results on NuScenes Val Set


What do the Learned Cluster Contexts Encode?


How do Cluster Contexts Affect the Behavior of Deformable Points?


MvACon's Computation Cost


Qualitative Results



Bibtex


	@InProceedings{Liu_2024_CVPR,
	    author    = {Liu, Xianpeng and Zheng, Ce and Qian, Ming and Xue, Nan and Chen, Chen and Zhang, Zhebin and Li, Chen and Wu, Tianfu},
	    title     = {Multi-View Attentive Contextualization for Multi-View 3D Object Detection},
	    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
	    month     = {June},
	    year      = {2024},
	    pages     = {16688-16698}
	}
        

This webpage template was adapted from here.