Method

ACF + LidarDepth [la] [ACFD]


Submitted on 30 May. 2016 11:22 by
Martin Dimitrievski (IPI/TELIN)

Running time:0.2 s
Environment:4 cores @ >3.5 Ghz (C/C++)

Method Description:
Parameters:
opts.modelDs=[50 20.5];
opts.modelDsPad=[64 32];
opts.nWeak=[64 256 1024 4096];
opts.nNeg=30000;
opts.nAccNeg=50000;
opts.pBoost.pTree.nThreads =16;
opts.stride = 2;
detector.opts.pPyramid.nPerOct = 8;
detector.opts.pPyramid.nApprox = 7;
detector.opts.pPyramid.nOctUp = 1;
Latex Bibtex:
@inproceedings{DBLP:conf/ivs/DimitrievskiVP17,
author = {Martin D. Dimitrievski and
Peter Veelaert and
Wilfried Philips},
title = {Semantically aware multilateral filter for depth upsampling in automotive
LiDAR point clouds},
booktitle = {{IEEE} Intelligent Vehicles Symposium, {IV} 2017, Los Angeles, CA,
USA, June 11-14, 2017},
pages = {1058--1063},
year = {2017},
crossref = {DBLP:conf/ivs/2017},
url = {https://doi.org/10.1109/IVS.2017.7995854},
doi = {10.1109/IVS.2017.7995854},
timestamp = {Sun, 06 Aug 2017 15:17:33 +0200},
biburl = {http://dblp.org/rec/bib/conf/ivs/DimitrievskiVP17},
bibsource = {dblp computer science bibliography, http://dblp.org}
}

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) 61.62 % 48.63 % 44.15 %
This table as LaTeX


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




eXTReMe Tracker