Method

Detection on Bird's Elevation Map [DoBEM]
[Anonymous Submission]

Submitted on 30 Aug. 2017 09:15 by
[Anonymous Submission]

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

Method Description:
Bird eye view based car detection using Faster R-CNN
Parameters:
init-net:VGG16
Latex Bibtex:
@INPROCEEDINGS {,
author = "Shang-Lin Yu and Thomas Westfechtel and Ryunosuke Hamada and Kazunori Ohno and Satoshi Tadokoro",
title = "Vehicle Detection and Localization on Bird's Eye View Elevation Images Using Convolutional Neural Network",
booktitle = "IEEE International Symposium on Safety, Security, and Rescue Robotics (SSRR)",
year = "2017"
}

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) 36.35 % 33.61 % 37.78 %
Car (Orientation) 15.35 % 14.02 % 16.33 %
Car (3D Detection) 7.42 % 6.95 % 13.45 %
Car (Bird's Eye View) 36.49 % 36.95 % 38.10 %
This table as LaTeX


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



Orientation estimation results.
This figure as: png eps pdf txt gnuplot



3D object detection results.
This figure as: png eps pdf txt gnuplot



Bird's eye view results.
This figure as: png eps pdf txt gnuplot




eXTReMe Tracker