The KITTI Vision Benchmark Suite

Method

Monocular 3D Object Detection with LiDAR Guided Semi Supervised Active Learning [MonoLiG]
https://arxiv.org/abs/2307.08415

Submitted on 1 Mar. 2023 16:24 by
Haluk Aral Hekimoglu (BMW Group)

Running time:		0.03 s
Environment:		1 core @ 2.5 Ghz (C/C++)

Method Description:

We train MonoFlex with additional labels from the

Kitti-depth dataset using LiDAR predictions as

ground-truth information similar. We filter LiDAR labels using aleatoric

uncertainty after showing theoratically that it

imposes additional errors from the LiDAR modality.

Parameters:

N/A

Latex Bibtex:

@misc{hekimoglu2023monocular,
title={Monocular 3D Object Detection with LiDAR Guided Semi
Supervised Active Learning},
author={Aral Hekimoglu and Michael Schmidt and Alvaro Marcos-
Ramiro},
year={2023},
eprint={2307.08415},
archivePrefix={arXiv},
primaryClass={cs.CV}
}

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)	96.70 %	93.56 %	83.74 %
Car (Orientation)	96.56 %	93.23 %	83.42 %
Car (3D Detection)	24.90 %	18.86 %	16.79 %
Car (Bird's Eye View)	35.73 %	26.83 %	24.24 %
Pedestrian (Detection)	62.87 %	47.69 %	43.27 %
Pedestrian (Orientation)	52.66 %	38.92 %	35.05 %
Pedestrian (3D Detection)	2.89 %	1.94 %	1.91 %
Pedestrian (Bird's Eye View)	3.74 %	2.72 %	2.55 %
Cyclist (Detection)	80.41 %	58.35 %	51.21 %
Cyclist (Orientation)	76.10 %	54.91 %	47.58 %
Cyclist (3D Detection)	8.14 %	5.24 %	4.45 %
Cyclist (Bird's Eye View)	9.48 %	6.49 %	5.46 %