Rethinking Voxelization and Classification for 3D Object Detection
This work addresses the problem of real-time and reliable 3D object detection for autonomous driving or robotics, but it appears incremental as it builds on existing pillar-based and voxel-based models.
The paper tackles the challenge of achieving real-time performance and reliability in 3D object detection from LiDAR point clouds by introducing a fast dynamic voxelizer and a lightweight detection sub-head, resulting in improved inference speed and precision with negligible time and computing cost.
The main challenge in 3D object detection from LiDAR point clouds is achieving real-time performance without affecting the reliability of the network. In other words, the detecting network must be confident enough about its predictions. In this paper, we present a solution to improve network inference speed and precision at the same time by implementing a fast dynamic voxelizer that works on fast pillar-based models in the same way a voxelizer works on slow voxel-based models. In addition, we propose a lightweight detection sub-head model for classifying predicted objects and filter out false detected objects that significantly improves model precision in a negligible time and computing cost. The developed code is publicly available at: https://github.com/YoushaaMurhij/RVCDet.