Method

Monocular 3D Object Detection for Autonomous Driving [Mono3D]
http://3dimage.ee.tsinghua.edu.cn/cxz/mono3d

Submitted on 27 Mar. 2016 03:30 by
Xiaozhi Chen (Tsinghua University)

Running time:4.2 s
Environment:GPU @ 2.5 Ghz (Matlab + C/C++)

Method Description:
Monocular 3D Object Detection for Autonomous Driving
Parameters:
Latex Bibtex:
@inproceedings{Chen2016CVPR,
title = {Monocular 3D Object Detection for Autonomous
Driving},
author = {Xiaozhi Chen and Kaustav Kundu and Ziyu Zhang
and Huimin Ma and Sanja Fidler and Raquel Urtasun},
booktitle = {CVPR},
year = {2016}
}

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) 94.52 % 89.37 % 79.15 %
Car (Orientation) 93.13 % 87.28 % 77.00 %
Pedestrian (Detection) 80.30 % 67.29 % 62.23 %
Pedestrian (Orientation) 71.19 % 58.66 % 53.94 %
Cyclist (Detection) 77.19 % 65.15 % 57.88 %
Cyclist (Orientation) 67.33 % 53.96 % 47.91 %
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



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



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



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



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




eXTReMe Tracker