We present a simple and flexible object detection framework optimized for autonomous driving. Building on the observation that point clouds in this application are extremely sparse, we propose a practical pillar-based approach to fix the imbalance issue caused by anchors. In particular, our algorithm incorporates a cylindrical projection into multi-view feature learning, predicts bounding box parameters per pillar rather than per point or per anchor, and includes an aligned pillar-to-point projection module to improve the final prediction. Our anchor-free approach avoids hyperparameter search associated with past methods, simplifying 3D object detection while significantly improving upon state-of-the-art.
@article{arxiv.2007.10323,
title = {Pillar-based Object Detection for Autonomous Driving},
author = {Yue Wang and Alireza Fathi and Abhijit Kundu and David Ross and Caroline Pantofaru and Thomas Funkhouser and Justin Solomon},
journal= {arXiv preprint arXiv:2007.10323},
year = {2020}
}