Method

Kinematic 3D Object Detection in Monocular Video [Kinematic3D]
http://cvlab.cse.msu.edu/project-kinematic.html

Submitted on 17 Jul. 2020 06:46 by
Garrick Brazil (Michigan State University)

Running time:0.12 s
Environment:1 core @ 1.5 Ghz (C/C++)

Method Description:
Kinematic 3D Object Detection in Monocular Video

By Garrick Brazil, Gerard Pons-Moll, Xiaoming Liu,
and Bernt Schiele
Parameters:
Please refer to paper and code for details.
Latex Bibtex:
@inproceedings{ brazil2020kinematic,
author = { Brazil, Garrick and Pons-Moll,
Gerard and Liu, Xiaoming and Schiele, Bernt},
title = { Kinematic 3D Object Detection in
Monocular Video},
booktitle = { ECCV },
address = { Virtual },
month = { August },
year = { 2020 },
organization = { Springer }
}

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) 89.67 % 71.73 % 54.97 %
Car (Orientation) 58.33 % 45.50 % 34.81 %
Car (3D Detection) 19.07 % 12.72 % 9.17 %
Car (Bird's Eye View) 26.69 % 17.52 % 13.10 %
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