Method

Regionlets Object Detector with Regionlets Re-localization [Regionlets]


Submitted on 14 Jan. 2015 00:03 by
Xiaoyu Wang (NEC LABS AMERICA)

Running time:1 s
Environment:>8 cores @ 2.5 Ghz (C/C++)

Method Description:
Regionlets detector based on multiple sources
object proposals. The detection results are
further refined by Regionlets Re-localization.


Details will be given in
http://www.xiaoyumu.com/project/kitti

This is an updated version of our results
submitted on Oct 6th which achieved:

Benchmark Easy Moderate Hard
Car (Detection) 84.27% 75.68% 59.20%
Pedestrian (Detection) 68.79% 55.01% 49.75%
Cyclist (Detection) 56.96% 44.65% 39.05%


Papers are at:

Regionlets + CNN:
http://www.xiaoyumu.com/s/PDF/Regionlets-pami.pdf
http://www.xiaoyumu.com/s/PDF/cnnRegionlet.pdf

Regionlets + Re-localization:

http://www.xiaoyumu.com/s/PDF/Regionlets_relocali
zation.pdf
Parameters:
No parameters
Latex Bibtex:
@inproceedings{Wang2015PAMI,
author = {Xiaoyu Wang and Ming Yang and Shenghuo
Zhu and Yuanqing Lin},
title = {Regionlets for Generic Object
Detection},
booktitle = {T-PAMI},
year = {2015},
}

@inproceedings{Zou2014BMVC,
author = {Will Y. Zou and Xiaoyu Wang and Miao
Sun and Yuanqing Lin},
title = {Generic Object Detection with Dense
Neural Patterns and Regionlets},
booktitle = {British Machine Vision Conference},
year = {2014},
}

@inproceedings{Regionlets-Relocalization,
author = {Chengjiang Long and Xiaoyu Wang and
Gang Hua and Ming Yang and Yuanqing Lin},
title = {Accurate Object Detection with Location
Relaxation and Regionlets Relocalization},
booktitle = {Asian Conference on Computer
Vision},
year = {2014},
}

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) 86.50 % 76.56 % 59.82 %
Pedestrian (Detection) 72.96 % 61.16 % 55.22 %
Cyclist (Detection) 70.09 % 58.69 % 51.81 %
This table as LaTeX


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



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



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




eXTReMe Tracker