Method

Probabilistic Graphical Models applied to Road Segmentation [PGM-ARS]


Submitted on 20 Jan. 2015 01:13 by
Mario Passani (University of Alcala)

Running time:0.05 s
Environment:i74700MQ @ 2.1Ghz (C/C++)

Method Description:
Machine learning approach based on a CRF to
incorporate some image descriptors
Parameters:
Latex Bibtex:
@inproceedings{Passani15IV,
author = {Mario Passani and J. Javier Yebes
and Luis M. Bergasa},
title = {Fast Pixelwise Road Inference Based on
Uniformly Reweighted Belief Propagation
},
booktitle = {Proc. {IEEE} Intelligent
Vehicles Symposium},
year = {2015},

}

Evaluation in Bird's Eye View


Benchmark MaxF AP PRE REC FPR FNR
UM_ROAD 80.97 % 69.11 % 77.51 % 84.76 % 11.21 % 15.24 %
UMM_ROAD 91.76 % 84.80 % 88.05 % 95.80 % 14.30 % 4.20 %
UU_ROAD 79.94 % 67.77 % 77.37 % 82.67 % 7.88 % 17.33 %
URBAN_ROAD 85.69 % 73.83 % 82.34 % 89.33 % 10.56 % 10.67 %
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.


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

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png


eXTReMe Tracker