Method

Pyramid R-CNN [Pyramid R-CNN]


Submitted on 4 Mar. 2021 12:42 by
Minzhe Niu (Shanghai Jiao Tong University)

Running time:0.07 s
Environment:1 core @ 2.5 Ghz (C/C++)

Method Description:
We propose RoI-grid Attention, density-aware radius
prediction, and pyramid aggregation module in our
model
Parameters:
\type = point cloud
Latex Bibtex:
@inproceedings{mao2021pyramid,
title={Pyramid R-CNN: Towards Better Performance and
Adaptability for 3D Object Detection},
author={Mao, Jiageng and Niu, Minzhe and Bai, Haoyue
and Liang, Xiaodan and Xu, Hang and Xu, Chunjing},
booktitle={ICCV},
year={2021}
}

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) 95.88 % 95.13 % 92.62 %
Car (Orientation) 95.87 % 95.03 % 92.46 %
Car (3D Detection) 88.39 % 82.08 % 77.49 %
Car (Bird's Eye View) 92.19 % 88.84 % 86.21 %
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