Aggregate Channel Features plus Semantic Context [ACF-SC]

Submitted on 1 Oct. 2014 12:22 by
Cesar Cadena Lerma (University of Adelaide)

Running time:<0.3 s
Environment:1 core @ >3.5 Ghz (Matlab + C/C++)

Method Description:
Re-scoring ACF detections by coarse semantic
context around the bounding box.
Baseline detector: same as ACF
Coverage for context: 1.5 x BB
Latex Bibtex:
title={A Fast, Modular Scene Understanding
System using Context-Aware Object Detection},
author={Cadena, C. and Dick, A. and Reid, I.},
booktitle = {Robotics and Automation (ICRA),
2015 IEEE International Conference on},
year = {2015},
address = {Seattle, WA, USA},
month = {May},
fullauthor = {Cadena, Cesar and Dick, Anthony
and Reid, Ian}

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) 69.76 % 55.76 % 46.27 %
Pedestrian (Detection) 54.20 % 44.77 % 39.57 %
This table as LaTeX

This figure as: png eps pdf txt gnuplot

This figure as: png eps pdf txt gnuplot

eXTReMe Tracker