The KITTI Vision Benchmark Suite

Method

USNet [USNet]
https://github.com/morancyc/USNet

Submitted on 15 Sep. 2021 11:39 by
yicong chang (Beijing University of Posts and Telecommunications)

Running time:		0.02 s
Environment:		GPU @ 1.5 Ghz (Python)

Method Description:

RGB-D road segmentation

Parameters:

TBA

Latex Bibtex:

@inproceedings{Chang22Fast,
author = {Chang, Yicong and
Xue, Feng and Sheng, Fei and
Liang, Wenteng and Ming, Anlong},
title = {Fast Road Segmentation via
Uncertainty-aware Symmetric Network},
booktitle={IEEE International Conference on
Robotics and Automation (ICRA)},
year={2022},
organization={IEEE}
}

Evaluation in Bird's Eye View

Benchmark	MaxF	AP	PRE	REC	FPR	FNR
UM_ROAD	96.46 %	92.78 %	96.32 %	96.60 %	1.68 %	3.40 %
UMM_ROAD	97.68 %	95.13 %	97.28 %	98.09 %	3.02 %	1.91 %
UU_ROAD	96.11 %	91.71 %	95.86 %	96.37 %	1.36 %	3.63 %
URBAN_ROAD	96.89 %	93.25 %	96.51 %	97.27 %	1.94 %	2.73 %

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