The KITTI Vision Benchmark Suite

Method

RoadFormer [RoadFormer]

Submitted on 19 Aug. 2023 08:43 by
Rui Fan (Tongji University)

Running time:		0.07 s
Environment:		GPU @ 2.5 Ghz (Python)

Method Description:

Freespace and road defect detection based on
Transformer.

Parameters:

tba

Latex Bibtex:

@ARTICLE{li2023roadformer,
title={RoadFormer: Duplex Transformer for
RGB-Normal Semantic Road Scene Parsing},
author={Jiahang Li and Yikang Zhang and Peng
Yun and Guangliang Zhou and Qijun Chen and Rui
Fan},
year={2024},
journal={IEEE Transactions on Intelligent
Vehicles},
}

Evaluation in Bird's Eye View

Benchmark	MaxF	AP	PRE	REC	FPR	FNR
UM_ROAD	97.02 %	93.34 %	96.84 %	97.20 %	1.45 %	2.80 %
UMM_ROAD	98.15 %	95.60 %	98.07 %	98.23 %	2.13 %	1.77 %
UU_ROAD	97.02 %	92.78 %	96.61 %	97.43 %	1.12 %	2.57 %
URBAN_ROAD	97.50 %	93.85 %	97.16 %	97.84 %	1.57 %	2.16 %

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