Method

Keypoint-3D [Keypoint-3D]


Submitted on 14 Oct. 2021 09:00 by
Dan Zhao (XIAN JIAOTONG UNIVERSITY)

Running time:14 s
Environment:1 core @ 2.5 Ghz (C/C++)

Method Description:
We present a monocular 3D object detection method that formulates the 3D object localization as paired keypoints regression problem.
Parameters:
none
Latex Bibtex:
none

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) 93.31 % 88.87 % 76.10 %
Car (3D Detection) 15.97 % 10.42 % 7.91 %
Car (Bird's Eye View) 23.16 % 15.54 % 11.83 %
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