The KITTI Vision Benchmark Suite

Method

Integrating LiDAR, Imagery, and Relative Depth for Enhanced Precision [UdeerLID]

Submitted on 12 Aug. 2024 09:58 by
ni tao (zhejiang university)

Running time:		0.01 s
Environment:		1 core @ 2.5 Ghz (C/C++)

Method Description:

An innovative approach to road segmentation that
integrates LiDAR point cloud data, visual imagery,
and relative depth maps derived from images.

Parameters:

lr = 1e-3

Latex Bibtex:

@misc{ni2024udeerlidintegratinglidarimage,
title={UdeerLID+: Integrating LiDAR, Image,
and Relative Depth with Semi-Supervised},
author={Tao Ni and Xin Zhan and Tao Luo and
Wenbin Liu and Zhan Shi and JunBo Chen},
year={2024},
eprint={2409.06197},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2409.06197},
}

Evaluation in Bird's Eye View

Benchmark	MaxF	AP	PRE	REC	FPR	FNR
UM_ROAD	96.94 %	93.62 %	97.09 %	96.79 %	1.32 %	3.21 %
UMM_ROAD	97.90 %	95.58 %	97.64 %	98.17 %	2.61 %	1.83 %
UU_ROAD	96.32 %	92.99 %	96.20 %	96.44 %	1.24 %	3.56 %
URBAN_ROAD	97.22 %	94.02 %	97.15 %	97.29 %	1.57 %	2.71 %

This table as LaTeX

Behavior Evaluation

Benchmark	PRE-20	F1-20	HR-20	PRE-30	F1-30	HR-30	PRE-40	F1-40	HR-40

This table as LaTeX

Road/Lane Detection

The following plots show precision/recall curves for the bird's eye view evaluation.

This figure as: png eps pdf

Distance-dependent Behavior Evaluation

The following plots show the F1 score/Precision/Hitrate with respect to the longitudinal distance which has been used for evaluation.

Visualization of Results

The following images illustrate the performance of the method qualitatively on a couple of test images. We first show results in the perspective image, followed by evaluation in bird's eye view. Here, red denotes false negatives, blue areas correspond to false positives and green represents true positives.

This figure as: png