DynStatF: An Efficient Feature Fusion Strategy for LiDAR 3D Object Detection

TL;DR

This paper introduces DynStaF, a novel feature fusion strategy that combines multi-frame semantic richness with current frame accuracy for improved LiDAR 3D object detection, achieving state-of-the-art results.

Contribution

The paper proposes DynStaF, a dual-path feature fusion method with Neighborhood Cross Attention and Dynamic-Static Interaction modules, enhancing multi-frame and single-frame feature integration.

Findings

Significant performance boost on nuScenes dataset, increasing PointPillars NDS from 57.7% to 61.6%.

Achieves state-of-the-art results with CenterPoint, reaching 61.0% mAP and 67.7% NDS.

Demonstrates effectiveness of dynamic-static feature fusion in LiDAR 3D detection.

Abstract

Augmenting LiDAR input with multiple previous frames provides richer semantic information and thus boosts performance in 3D object detection, However, crowded point clouds in multi-frames can hurt the precise position information due to the motion blur and inaccurate point projection. In this work, we propose a novel feature fusion strategy, DynStaF (Dynamic-Static Fusion), which enhances the rich semantic information provided by the multi-frame (dynamic branch) with the accurate location information from the current single-frame (static branch). To effectively extract and aggregate complimentary features, DynStaF contains two modules, Neighborhood Cross Attention (NCA) and Dynamic-Static Interaction (DSI), operating through a dual pathway architecture. NCA takes the features in the static branch as queries and the features in the dynamic branch as keys (values). When computing the attention, we address the sparsity of point clouds and take only neighborhood positions into consideration. NCA fuses two features at different feature map scales, followed by DSI providing the comprehensive interaction. To analyze our proposed strategy DynStaF, we conduct extensive experiments on the nuScenes dataset. On the test set, DynStaF increases the performance of PointPillars in NDS by a large margin from 57.7% to 61.6%. When combined with CenterPoint, our framework achieves 61.0% mAP and 67.7% NDS, leading to state-of-the-art performance without bells and whistles.