The KITTI Vision Benchmark Suite

Method

MonOri: Orientation-Guided PnP for Monocular 3-D Object Detection [MonOri]
https://ieeexplore.ieee.org/document/11054071

Submitted on 30 Jun. 2025 17:17 by
Hongdou Yao (Wuhan University)

Running time:		0.03 s
Environment:		4 cores @ 2.5 Ghz (Python)

Method Description:

MonOri: Orientation-Guided PnP for Monocular 3-D
Object Detection

Parameters:

learning_rate = 2e − 4,
weight_decay = 1e − 5,

Latex Bibtex:

@ARTICLE{11054071,
author={Yao, Hongdou and Han, Pengfei and Chen,
Jun and Wang, Zheng and Qiu, Yansheng and Wang, Xiao
and wang, Yimin and Chai, Xiaoyu and Cao, Chenglong
and Jin, Wei},
journal={IEEE Transactions on Neural Networks and
Learning Systems},
title={MonOri: Orientation-Guided PnP for
Monocular 3-D Object Detection},
year={2025},
volume={},
number={},
pages={1-13},
keywords={3-D object detection;keypoints’
selection;monocular image;occlusion
object;orientation guided},
doi={10.1109/TNNLS.2025.3577618}}

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)	95.23 %	88.72 %	81.77 %
Car (Orientation)	95.19 %	88.58 %	81.58 %
Car (3D Detection)	25.20 %	16.77 %	14.45 %
Car (Bird's Eye View)	37.86 %	26.00 %	23.10 %
Pedestrian (Detection)	61.46 %	48.07 %	42.22 %
Pedestrian (Orientation)	56.28 %	43.12 %	37.80 %
Pedestrian (3D Detection)	18.97 %	12.76 %	11.00 %
Pedestrian (Bird's Eye View)	20.86 %	14.35 %	12.45 %
Cyclist (Detection)	71.12 %	56.05 %	48.34 %
Cyclist (Orientation)	65.05 %	50.13 %	43.51 %
Cyclist (3D Detection)	9.47 %	5.87 %	5.35 %
Cyclist (Bird's Eye View)	12.34 %	7.86 %	7.05 %