Road scene segmentation from a single image

Jose M. Alvarez, Theo Gevers, Yann LeCun, Antonio M. Lopez

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Road scene segmentation is important in computer vision for different applications such as autonomous driving and pedestrian detection. Recovering the 3D structure of road scenes provides relevant contextual information to improve their understanding. In this paper, we use a convolutional neural network based algorithm to learn features from noisy labels to recover the 3D scene layout of a road image. The novelty of the algorithm relies on generating training labels by applying an algorithm trained on a general image dataset to classify on-board images. Further, we propose a novel texture descriptor based on a learned color plane fusion to obtain maximal uniformity in road areas. Finally, acquired (off-line) and current (on-line) information are combined to detect road areas in single images. From quantitative and qualitative experiments, conducted on publicly available datasets, it is concluded that convolutional neural networks are suitable for learning 3D scene layout from noisy labels and provides a relative improvement of 7% compared to the baseline. Furthermore, combining color planes provides a statistical description of road areas that exhibits maximal uniformity and provides a relative improvement of 8% compared to the baseline. Finally, the improvement is even bigger when acquired and current information from a single image are combined.

Original languageEnglish (US)
Title of host publicationComputer Vision, ECCV 2012 - 12th European Conference on Computer Vision, Proceedings
Pages376-389
Number of pages14
Volume7578 LNCS
EditionPART 7
DOIs
StatePublished - 2012
Event12th European Conference on Computer Vision, ECCV 2012 - Florence, Italy
Duration: Oct 7 2012Oct 13 2012

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 7
Volume7578 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other12th European Conference on Computer Vision, ECCV 2012
CountryItaly
CityFlorence
Period10/7/1210/13/12

Fingerprint

Labels
Segmentation
Color
Neural networks
Uniformity
Layout
Baseline
Computer vision
Neural Networks
Pedestrian Detection
Fusion reactions
Textures
Computer Vision
Descriptors
Texture
Fusion
Classify
Experiments
Line
Experiment

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Alvarez, J. M., Gevers, T., LeCun, Y., & Lopez, A. M. (2012). Road scene segmentation from a single image. In Computer Vision, ECCV 2012 - 12th European Conference on Computer Vision, Proceedings (PART 7 ed., Vol. 7578 LNCS, pp. 376-389). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7578 LNCS, No. PART 7). https://doi.org/10.1007/978-3-642-33786-4_28

Road scene segmentation from a single image. / Alvarez, Jose M.; Gevers, Theo; LeCun, Yann; Lopez, Antonio M.

Computer Vision, ECCV 2012 - 12th European Conference on Computer Vision, Proceedings. Vol. 7578 LNCS PART 7. ed. 2012. p. 376-389 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7578 LNCS, No. PART 7).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Alvarez, JM, Gevers, T, LeCun, Y & Lopez, AM 2012, Road scene segmentation from a single image. in Computer Vision, ECCV 2012 - 12th European Conference on Computer Vision, Proceedings. PART 7 edn, vol. 7578 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 7, vol. 7578 LNCS, pp. 376-389, 12th European Conference on Computer Vision, ECCV 2012, Florence, Italy, 10/7/12. https://doi.org/10.1007/978-3-642-33786-4_28
Alvarez JM, Gevers T, LeCun Y, Lopez AM. Road scene segmentation from a single image. In Computer Vision, ECCV 2012 - 12th European Conference on Computer Vision, Proceedings. PART 7 ed. Vol. 7578 LNCS. 2012. p. 376-389. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 7). https://doi.org/10.1007/978-3-642-33786-4_28
Alvarez, Jose M. ; Gevers, Theo ; LeCun, Yann ; Lopez, Antonio M. / Road scene segmentation from a single image. Computer Vision, ECCV 2012 - 12th European Conference on Computer Vision, Proceedings. Vol. 7578 LNCS PART 7. ed. 2012. pp. 376-389 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 7).
@inproceedings{0bbe58ad63a04cafbe95e224c7824ad3,
title = "Road scene segmentation from a single image",
abstract = "Road scene segmentation is important in computer vision for different applications such as autonomous driving and pedestrian detection. Recovering the 3D structure of road scenes provides relevant contextual information to improve their understanding. In this paper, we use a convolutional neural network based algorithm to learn features from noisy labels to recover the 3D scene layout of a road image. The novelty of the algorithm relies on generating training labels by applying an algorithm trained on a general image dataset to classify on-board images. Further, we propose a novel texture descriptor based on a learned color plane fusion to obtain maximal uniformity in road areas. Finally, acquired (off-line) and current (on-line) information are combined to detect road areas in single images. From quantitative and qualitative experiments, conducted on publicly available datasets, it is concluded that convolutional neural networks are suitable for learning 3D scene layout from noisy labels and provides a relative improvement of 7{\%} compared to the baseline. Furthermore, combining color planes provides a statistical description of road areas that exhibits maximal uniformity and provides a relative improvement of 8{\%} compared to the baseline. Finally, the improvement is even bigger when acquired and current information from a single image are combined.",
author = "Alvarez, {Jose M.} and Theo Gevers and Yann LeCun and Lopez, {Antonio M.}",
year = "2012",
doi = "10.1007/978-3-642-33786-4_28",
language = "English (US)",
isbn = "9783642337857",
volume = "7578 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
number = "PART 7",
pages = "376--389",
booktitle = "Computer Vision, ECCV 2012 - 12th European Conference on Computer Vision, Proceedings",
edition = "PART 7",

}

TY - GEN

T1 - Road scene segmentation from a single image

AU - Alvarez, Jose M.

AU - Gevers, Theo

AU - LeCun, Yann

AU - Lopez, Antonio M.

PY - 2012

Y1 - 2012

N2 - Road scene segmentation is important in computer vision for different applications such as autonomous driving and pedestrian detection. Recovering the 3D structure of road scenes provides relevant contextual information to improve their understanding. In this paper, we use a convolutional neural network based algorithm to learn features from noisy labels to recover the 3D scene layout of a road image. The novelty of the algorithm relies on generating training labels by applying an algorithm trained on a general image dataset to classify on-board images. Further, we propose a novel texture descriptor based on a learned color plane fusion to obtain maximal uniformity in road areas. Finally, acquired (off-line) and current (on-line) information are combined to detect road areas in single images. From quantitative and qualitative experiments, conducted on publicly available datasets, it is concluded that convolutional neural networks are suitable for learning 3D scene layout from noisy labels and provides a relative improvement of 7% compared to the baseline. Furthermore, combining color planes provides a statistical description of road areas that exhibits maximal uniformity and provides a relative improvement of 8% compared to the baseline. Finally, the improvement is even bigger when acquired and current information from a single image are combined.

AB - Road scene segmentation is important in computer vision for different applications such as autonomous driving and pedestrian detection. Recovering the 3D structure of road scenes provides relevant contextual information to improve their understanding. In this paper, we use a convolutional neural network based algorithm to learn features from noisy labels to recover the 3D scene layout of a road image. The novelty of the algorithm relies on generating training labels by applying an algorithm trained on a general image dataset to classify on-board images. Further, we propose a novel texture descriptor based on a learned color plane fusion to obtain maximal uniformity in road areas. Finally, acquired (off-line) and current (on-line) information are combined to detect road areas in single images. From quantitative and qualitative experiments, conducted on publicly available datasets, it is concluded that convolutional neural networks are suitable for learning 3D scene layout from noisy labels and provides a relative improvement of 7% compared to the baseline. Furthermore, combining color planes provides a statistical description of road areas that exhibits maximal uniformity and provides a relative improvement of 8% compared to the baseline. Finally, the improvement is even bigger when acquired and current information from a single image are combined.

UR - http://www.scopus.com/inward/record.url?scp=84867871457&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84867871457&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-33786-4_28

DO - 10.1007/978-3-642-33786-4_28

M3 - Conference contribution

AN - SCOPUS:84867871457

SN - 9783642337857

VL - 7578 LNCS

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 376

EP - 389

BT - Computer Vision, ECCV 2012 - 12th European Conference on Computer Vision, Proceedings

ER -