Method

Learning Object Sub-Categories [SubCat]


Submitted on 13 Jul. 2014 17:38 by
Eshed Ohn-Bar (UCSD)

Running time:1.2 s
Environment:6 cores @ 2.5 Ghz (Matlab + C/C++)

Method Description:
Cluster to visual (K=20) and geometric (16
orientation bins) subcategories.
Parameters:
-
Latex Bibtex:
@InProceedings{Ohn-Bar2014CVPRWORK,
author = {Eshed Ohn-Bar and Mohan M. Trivedi},
title = {Fast and Robust Object Detection Using
Visual Subcategories},
booktitle = {Computer Vision and Pattern
Recognition
Workshops Mobile 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
Pedestrian (Detection) 53.75 % 40.50 % 35.66 %
Pedestrian (Orientation) 42.31 % 31.26 % 27.39 %
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