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

TL;DR

This paper introduces MvACon, a novel attentive contextualization method that enhances multi-view 3D object detection by effectively utilizing dense scene-level features, leading to improved accuracy across multiple benchmarks.

Contribution

MvACon provides a dense yet computationally efficient attentive contextualization scheme that improves 2D-to-3D feature lifting in multi-view 3D detection, agnostic to specific lifting approaches.

Findings

Consistent performance improvements on nuScenes benchmark.

Enhanced detection accuracy in location, orientation, and velocity.

Effective encoding of dense scene-level contexts.

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".