Method

RefineNet: Iterative Refinement for Accurate Object Localization [RefineNet]


Submitted on 23 May. 2016 23:06 by
Rakesh Nattoji Rajaram (CVRR-UCSD)

Running time:0.20 s
Environment:GPU @ 2.5 Ghz (Matlab + C++)

Method Description:
Faster R-CNN + Iterative bbs refinement using ZF
network.
Parameters:
Latex Bibtex:
@article{7944662,
author={Rakesh Nattoji Rajaram and Eshed Ohn-
Bar and Mohan M. Trivedi},
journal={IEEE Transactions on Intelligent
Vehicles},
title={RefineNet: Refining Object Detectors for
Autonomous Driving},
year={2016},
month={Dec},}

@article{Rajaram2016ITSC,
author = {Rakesh Nattoji Rajaram and Eshed Ohn-
Bar and Mohan M. Trivedi},
title = {RefineNet: Iterative Refinement for
Accurate Object Localization},
journal = {Intelligent Transportation Systems
Conference},
month = {Nov},
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) 91.91 % 81.01 % 65.67 %
This table as LaTeX

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




eXTReMe Tracker