Method

RangeRCNN: Towards Fast and Accurate 3D Object Detection with Range Image Representation [la] [RangeRCNN]


Submitted on 2 Sep. 2020 04:01 by
Zhidong Liang (Hikvision Research Institute)

Running time:0.06 s
Environment:GPU @ 2.5 Ghz (Python + C/C++)

Method Description:
See the paper.
Parameters:
See the paper.
Latex Bibtex:
@article{liang2020rangercnn,
title={RangeRCNN: Towards Fast and Accurate 3D
Object Detection with Range Image
Representation},
author={Liang, Zhidong and Zhang, Ming and
Zhang, Zehan and Zhao, Xian and Pu, Shiliang},
journal={arXiv preprint arXiv:2009.00206},
year={2020}
}

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 (Detection) 95.48 % 94.03 % 91.74 %
Car (Orientation) 95.47 % 93.90 % 91.53 %
Car (3D Detection) 88.47 % 81.33 % 77.09 %
Car (Bird's Eye View) 92.15 % 88.40 % 85.74 %
This table as LaTeX


2D object detection results.
This figure as: png eps pdf txt gnuplot



Orientation estimation results.
This figure as: png eps pdf txt gnuplot



3D object detection results.
This figure as: png eps pdf txt gnuplot



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




eXTReMe Tracker