Many point-based 3D detectors adopt point-feature sampling strategies to drop some points for efficient inference. These strategies are typically based on fixed and handcrafted rules, making it difficult to handle complicated scenes. Different from them, we propose a Dynamic Ball Query (DBQ) network to adaptively select a subset of input points according to the input features, and assign the feature transform with a suitable receptive field for each selected point. It can be embedded into some state-of-the-art 3D detectors and trained in an end-to-end manner, which significantly reduces the computational cost. Extensive experiments demonstrate that our method can increase the inference speed by 30%-100% on KITTI, Waymo, and ONCE datasets. Specifically, the inference speed of our detector can reach 162 FPS on KITTI scene, and 30 FPS on Waymo and ONCE scenes without performance degradation. Due to skipping the redundant points, some evaluation metrics show significant improvements. Codes will be released at https://github.com/yancie-yjr/DBQ-SSD.
@article{arxiv.2207.10909,
title = {DBQ-SSD: Dynamic Ball Query for Efficient 3D Object Detection},
author = {Jinrong Yang and Lin Song and Songtao Liu and Weixin Mao and Zeming Li and Xiaoping Li and Hongbin Sun and Jian Sun and Nanning Zheng},
journal= {arXiv preprint arXiv:2207.10909},
year = {2023}
}
Comments
ICLR 2023. Code at https://github.com/yancie-yjr/DBQ-SSD