Method

PIXOR - LiDAR Only [la] [UberATG-PIXOR]


Submitted on 8 Aug. 2018 13:26 by
Bin Yang (University of Toronto)

Running time:0.035 s
Environment:TITAN Xp (Python)

Method Description:
Customized ResNet trained from scratch on KITTI
LIDAR training data only.
Fixed the wrong timing in the paper.
Parameters:
See the paper.
Latex Bibtex:
@inproceedings{Yang2018CVPR,
title = {PIXOR: Real-time 3D Object Detection from
Point
Clouds},
author = {Bin Yang and Wenjie Luo and Raquel
Urtasun},
booktitle = {CVPR},
year = {2018}
}

Detailed Results

Object detection and orientation estimation results. Results for object detection are given in terms of average precision (AP) and results for joint object detection and orientation estimation are provided in terms of average orientation similarity (AOS).


Benchmark Easy Moderate Hard
Car (Bird's Eye View) 83.97 % 80.01 % 74.31 %
This table as LaTeX


Bird's eye view results.
This figure as: png eps pdf txt gnuplot




eXTReMe Tracker