Method

Scale Dependent Pooling with Box proposals [SDP+CRC (ft)]


Submitted on 6 Nov. 2015 16:05 by
Wongun Choi (NEC Laboratories)

Running time:0.6 s
Environment:GPU @ 2.5 Ghz (C/C++)

Method Description:
Please see the paper for details.
Parameters:
Please see the paper for details.
Latex Bibtex:
@inproceedings{Yang2016CVPR,
title={Exploit All the Layers: Fast and Accurate CNN
Object Detector with Scale Dependent Pooling and
Cascaded Rejection Classifiers},
author={Fan Yang and Wongun Choi and Yuanqing Lin},
booktitle={Proceedings of the IEEE International
Conference on Computer Vision and Pattern Recognition},
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) 92.06 % 85.00 % 71.71 %
Pedestrian (Detection) 79.22 % 64.36 % 59.16 %
Cyclist (Detection) 75.63 % 60.72 % 53.00 %
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