Method

hybrid cascade + maxFtr + ROI [maxFtr+ROI]
[Anonymous Submission]

Submitted on 29 Aug. 2016 18:08 by
[Anonymous Submission]

Running time:0.25 s
Environment:4 cores @ 2.5 Ghz (C/C++)

Method Description:
a hybrid cascade consisting decision trees and support vector machine plus max pooled features and ROI from geometry constraint.
Parameters:
3-4 stages
Latex Bibtex:
@inproceedings{Tian2017VISAPP,
title={Detection and Orientation Estimation for Cyclists by Max Pooled
Features},
author={Tian, Wei and Lauer, Martin},
booktitle={International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP)},
year={2017},
}

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
Cyclist (Detection) 49.65 % 43.59 % 38.74 %
Cyclist (Orientation) 42.96 % 38.29 % 34.28 %
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




eXTReMe Tracker