Method

Efficient fine-grained road segmentation using superpixel-based CNN and CRF models [FRS_SP]
[Anonymous Submission]

Submitted on 16 Oct. 2017 14:18 by
[Anonymous Submission]

Running time:0.21 s
Environment:4 cores @ 3.0 Ghz (Matlab + C/C++)

Method Description:
road segmentation based on CNN and crf techniques
Parameters:
learning rate=0.1*10^-4
weight decay=0.0001
Momentum=0.9
mini-batches size= 50
CRF potentioal=0.1
Latex Bibtex:

Evaluation in Bird's Eye View


Benchmark MaxF AP PRE REC FPR FNR
UM_ROAD 83.22 % 72.94 % 77.11 % 90.39 % 12.23 % 9.61 %
UMM_ROAD 90.96 % 84.63 % 87.86 % 94.29 % 14.32 % 5.71 %
UU_ROAD 80.02 % 67.93 % 77.56 % 82.64 % 7.79 % 17.36 %
URBAN_ROAD 85.97 % 77.81 % 82.04 % 90.31 % 10.89 % 9.69 %
This table as LaTeX

Behavior Evaluation


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