The KITTI Vision Benchmark Suite

Method

Fully Convolutional Neural Network with Large Context [FCN-LC]

Submitted on 27 Aug. 2015 17:44 by
Caio César Teodoro Mendes (University of São Paulo)

Running time:		0.03 s
Environment:		GPU Titan X

Method Description:

Convolutional ANN trained on patches and inference
done using a fully convolutional network.

Video demo: https://youtu.be/_VESRS81Pvs

Parameters:

conv 32

Latex Bibtex:

@INPROCEEDINGS{Mendes2016ICRA,
author={Caio César Teodoro Mendes and Vincent
Frémont and Denis Fernando Wolf},
booktitle={IEEE Conference on Robotics and
Automation (ICRA)},
title={Exploiting Fully Convolutional Neural
Networks for Fast Road Detection},
year={2016},
month={May}
}

Evaluation in Bird's Eye View

Benchmark	MaxF	AP	PRE	REC	FPR	FNR
UM_ROAD	89.36 %	78.80 %	89.35 %	89.37 %	4.85 %	10.63 %
UMM_ROAD	94.09 %	90.26 %	94.05 %	94.13 %	6.55 %	5.87 %
UU_ROAD	86.27 %	75.37 %	86.65 %	85.89 %	4.31 %	14.11 %
URBAN_ROAD	90.79 %	85.83 %	90.87 %	90.72 %	5.02 %	9.28 %

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