Method

Multi-View Adaptive Fusion Network for 3D Object Detection [MVAF-Net]
https://github.com/wangguojun2018/MVAF-Net

Submitted on 5 Dec. 2020 07:18 by
Guoyun Wang (the Baidu Inc.)

Running time:0.06 s
Environment:1 core @ 2.5 Ghz (Python + C/C++)

Method Description:
TBD
Parameters:
TBD
Latex Bibtex:
@article{wang2020multi,
title={Multi-View Adaptive Fusion Network for
3D Object Detection},
author={Wang, Guojun and Tian, Bin and Zhang,
Yachen and Chen, Long and Cao, Dongpu and Wu,
Jian},
journal={arXiv preprint arXiv:2011.00652},
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) 95.37 % 93.66 % 90.90 %
Car (Orientation) 95.35 % 93.54 % 90.70 %
Car (3D Detection) 87.87 % 78.71 % 75.48 %
Car (Bird's Eye View) 91.95 % 87.73 % 85.00 %
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