Method

AlexNet over SquaresChnFtrs proposals [R-CNN]


Submitted on 26 Jan. 2015 18:12 by
Rodrigo Benenson (MPI-Inf)

Running time:4 s
Environment:GPU @ 3.3 Ghz (C/C++)

Method Description:
R-CNN evaluated over SquaresChnFtrs proposals
(named SquaresICF in KITTI benchmark).
R-CNN is pretrained on ImageNet and fine-tuned over
KITTI training set.
SquaresChnFtrs is trained on KITTI training set.

Further details in: http://arxiv.org/abs/1501.05790
Parameters:
See paper for details.
Latex Bibtex:
@INPROCEEDINGS{Hosang2015DnnForPedestrians,
author = {Jan Hosang and Mohamed Omran and
Rodrigo Benenson and Bernt
Schiele},
title = {Taking a Deeper Look at Pedestrians},
booktitle = {arXiv},
year = {2015}
}

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) 62.88 % 48.57 % 43.05 %
This table as LaTeX


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




eXTReMe Tracker