Method

A3DODWTDA [la] [A3DODWTDA]
https://github.com/fregu856/3DOD_thesis

Submitted on 20 Apr. 2018 10:47 by
Fredrik Gustafsson (Linköping University)

Running time:0.08 s
Environment:GPU @ 3.0 Ghz (Python)

Method Description:
Master's thesis project, see thesis for details:
http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-
148585
Parameters:
-
Latex Bibtex:
@mastersthesis{erino397fregu856master2018,
author = {Gustafsson, Fredrik and Linder-Norén,
Erik},
title = {Automotive 3D Object Detection Without
Target Domain Annotations},
school = {Linköping University},
year = 2018,
}

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) 82.98 % 79.15 % 68.30 %
Car (3D Detection) 62.84 % 56.82 % 48.12 %
Car (Bird's Eye View) 79.58 % 73.26 % 62.77 %
This table as LaTeX

2D object detection 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