Attention-based Proposals Refinement for 3D Object Detection

TL;DR

This paper introduces APRO3D-Net, a data-driven 3D object detection refinement method using Vector Attention, achieving high accuracy with fewer parameters and real-time speed on benchmark datasets.

Contribution

It proposes a novel Vector Attention-based refinement stage for 3D detection, reducing manual tuning and improving relation modeling between pooled points and ROI.

Findings

Achieves 84.85 AP for Car on KITTI at moderate difficulty.

Attains 47.03 mAP on NuScenes across 10 classes.

Operates at 15 FPS with fewer parameters.

Abstract

Recent advances in 3D object detection are made by developing the refinement stage for voxel-based Region Proposal Networks (RPN) to better strike the balance between accuracy and efficiency. A popular approach among state-of-the-art frameworks is to divide proposals, or Regions of Interest (ROI), into grids and extract features for each grid location before synthesizing them to form ROI features. While achieving impressive performances, such an approach involves several hand-crafted components (e.g. grid sampling, set abstraction) which requires expert knowledge to be tuned correctly. This paper proposes a data-driven approach to ROI feature computing named APRO3D-Net which consists of a voxel-based RPN and a refinement stage made of Vector Attention. Unlike the original multi-head attention, Vector Attention assigns different weights to different channels within a point feature, thus being able to capture a more sophisticated relation between pooled points and ROI. Our method achieves a competitive performance of 84.85 AP for class Car at moderate difficulty on the validation set of KITTI and 47.03 mAP (average over 10 classes) on NuScenes while having the least parameters compared to closely related methods and attaining an inference speed at 15 FPS on NVIDIA V100 GPU. The code is released at https://github.com/quan-dao/APRO3D-Net.