Ada3D : Exploiting the Spatial Redundancy with Adaptive Inference for Efficient 3D Object Detection

TL;DR

Ada3D introduces an adaptive inference framework that significantly reduces computational and memory costs in 3D object detection by filtering redundant input data based on spatial importance, without losing accuracy.

Contribution

The paper presents Ada3D, a novel adaptive inference method that exploits spatial redundancy in Lidar point clouds to improve efficiency in 3D object detection.

Findings

Achieves 40% reduction in 3D voxel data.

Reduces BEV feature map density from 100% to 20%.

Cuts model computational and memory costs by 5x.

Abstract

Voxel-based methods have achieved state-of-the-art performance for 3D object detection in autonomous driving. However, their significant computational and memory costs pose a challenge for their application to resource-constrained vehicles. One reason for this high resource consumption is the presence of a large number of redundant background points in Lidar point clouds, resulting in spatial redundancy in both 3D voxel and dense BEV map representations. To address this issue, we propose an adaptive inference framework called Ada3D, which focuses on exploiting the input-level spatial redundancy. Ada3D adaptively filters the redundant input, guided by a lightweight importance predictor and the unique properties of the Lidar point cloud. Additionally, we utilize the BEV features' intrinsic sparsity by introducing the Sparsity Preserving Batch Normalization. With Ada3D, we achieve 40% reduction for 3D voxels and decrease the density of 2D BEV feature maps from 100% to 20% without sacrificing accuracy. Ada3D reduces the model computational and memory cost by 5x, and achieves 1.52x/1.45x end-to-end GPU latency and 1.5x/4.5x GPU peak memory optimization for the 3D and 2D backbone respectively.