Method

GraR-Po [GraR-Po]
https://github.com/Nightmare-n/GraphRCNN

Submitted on 26 Dec. 2021 03:07 by
Honghui Yang (Zhejiang University)

Running time:0.06 s
Environment:1 core @ 2.5 Ghz (Python + C/C++)

Method Description:
TBD
Parameters:
TBD
Latex Bibtex:
@inproceedings{yang2022graphrcnn,
author = {Honghui Yang and
Zili Liu and
Xiaopei Wu and
Wenxiao Wang and
Wei Qian and
Xiaofei He and
Deng Cai},
title = {Graph R-CNN: Towards Accurate
3D Object Detection with Semantic-Decorated Local
Graph},
booktitle = {ECCV},
year = {2022},
}

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) 96.84 % 96.18 % 91.11 %
Car (Orientation) 96.83 % 96.09 % 90.99 %
Car (3D Detection) 91.79 % 83.18 % 77.98 %
Car (Bird's Eye View) 95.79 % 92.12 % 87.11 %
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