Method

MDSNet [MDSNet]


Submitted on 21 Feb. 2022 06:20 by
zhouzhen xie (zhejiang university)

Running time:0.05 s
Environment:1 core @ 2.5 Ghz (Python)

Method Description:
TBD
Parameters:
TBD
Latex Bibtex:
@article{xie2022mds,
title={MDS-Net: Multi-Scale Depth Stratification
3D Object Detection from Monocular Images},
author={Xie, Zhouzhen and Song, Yuying and Wu,
Jingxuan and Li, Zecheng and Song, Chunyi and Xu,
Zhiwei},
journal={Sensors},
volume={22},
number={16},
pages={6197},
year={2022},
publisher={MDPI}
}

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) 85.94 % 62.74 % 50.27 %
Car (3D Detection) 24.30 % 14.46 % 11.12 %
Car (Bird's Eye View) 32.81 % 20.14 % 15.77 %
Pedestrian (Detection) 41.64 % 29.25 % 26.01 %
Pedestrian (3D Detection) 10.68 % 7.09 % 6.06 %
Pedestrian (Bird's Eye View) 12.05 % 8.18 % 7.03 %
Cyclist (Detection) 28.23 % 16.64 % 14.14 %
Cyclist (3D Detection) 5.37 % 2.68 % 2.22 %
Cyclist (Bird's Eye View) 5.99 % 3.22 % 2.62 %
This table as LaTeX


2D object detection 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



2D object detection 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



2D object detection 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