Method

Geometry Uncertainty Projection Network for Monocular 3D Object Detection [GUPNet]
https://github.com/SuperMHP/GUPNet

Submitted on 2 Oct. 2021 12:04 by
Yan Lu (University of Sydney)

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

Method Description:
We use geometry to guide uncertainty learning and
achieve the new state-of-the-art performance on
KITTI about monocular 3D detection. This work has
been accepted by ICCV2021
Parameters:
\alpha=0.2
Latex Bibtex:
@article{lu2021geometry,
title={Geometry Uncertainty Projection Network
for Monocular 3D Object Detection},
author={Lu, Yan and Ma, Xinzhu and Yang, Lei and
Zhang, Tianzhu and Liu, Yating and Chu, Qi and Yan,
Junjie and Ouyang, Wanli},
journal={arXiv preprint arXiv:2107.13774},
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) 94.15 % 86.45 % 74.18 %
Car (Orientation) 93.92 % 85.90 % 73.55 %
Car (3D Detection) 22.26 % 15.02 % 13.12 %
Car (Bird's Eye View) 30.29 % 21.19 % 18.20 %
Pedestrian (Detection) 74.95 % 55.65 % 48.44 %
Pedestrian (Orientation) 68.93 % 50.74 % 44.01 %
Pedestrian (3D Detection) 14.95 % 9.76 % 8.41 %
Pedestrian (Bird's Eye View) 15.62 % 10.37 % 8.79 %
Cyclist (Detection) 67.11 % 42.78 % 37.94 %
Cyclist (Orientation) 55.03 % 35.03 % 31.18 %
Cyclist (3D Detection) 5.58 % 3.21 % 2.66 %
Cyclist (Bird's Eye View) 6.94 % 3.85 % 3.64 %
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



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



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