PIXOR: Real-time 3D Object Detection from Point Clouds

TLDR

Speed is critical for safety, yet existing 3D point‑cloud detection methods are computationally expensive due to high dimensionality. The study proposes PIXOR, a proposal‑free, single‑stage detector that efficiently processes point clouds from a Bird's Eye View to enable real‑time 3D object detection for autonomous driving. PIXOR uses a BEV input representation, a tailored network architecture, and optimized training to balance accuracy and speed, and is evaluated on the KITTI BEV benchmark and a large‑scale 3D vehicle detection dataset. On both benchmarks, PIXOR outperforms state‑of‑the‑art methods in Average Precision while maintaining 10 FPS real‑time performance.

Abstract

We address the problem of real-time 3D object detection from point clouds in the context of autonomous driving. Speed is critical as detection is a necessary component for safety. Existing approaches are, however, expensive in computation due to high dimensionality of point clouds. We utilize the 3D data more efficiently by representing the scene from the Bird's Eye View (BEV), and propose PIXOR, a proposal-free, single-stage detector that outputs oriented 3D object estimates decoded from pixel-wise neural network predictions. The input representation, network architecture, and model optimization are specially designed to balance high accuracy and real-time efficiency. We validate PIXOR on two datasets: the KITTI BEV object detection benchmark, and a large-scale 3D vehicle detection benchmark. In both datasets we show that the proposed detector surpasses other state-of-the-art methods notably in terms of Average Precision (AP), while still runs at 10 FPS.