Method

Weakly Supervised 3D Object Detection from Lidar Point Cloud[la] [WS3D]


Submitted on 28 Feb. 2020 15:39 by
Qinghao Meng (Beijing Institute of Technology)

Running time:0.1 s
Environment: GPU @ 2.5 Ghz (Python)

Method Description:
TBD.
Parameters:
TBD.
Latex Bibtex:
@inproceedings{meng_2020_eccv,
author={Meng, Qinghao and Wang, Wenguan and
Zhou, Tianfei and Shen, Jianbing and Van Gool,
Luc and Dai, Dengxin},
title ={Weakly Supervised 3D Object Detection
from Lidar Point Cloud},
url={https://arxiv.org/abs/2007.11901v1},
booktitle=ECCV,
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) 95.13 % 91.15 % 86.52 %
Car (Orientation) 94.85 % 90.69 % 85.94 %
Car (3D Detection) 80.99 % 70.59 % 64.23 %
Car (Bird's Eye View) 90.96 % 84.93 % 77.96 %
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