A convolutional neural network for pavement surface crack segmentation using residual connections and attention gating

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

6 Downloads (Pure)

Abstract

Conventional surface crack segmentation requires images manually labelled by a trained expert. It is a challenging task as cracks can vary in orientation and size, with some parts of cracks only being one pixel wide. Further, available training data for crack segmentation is sparse. In this work we propose to automate this annotation task, by introducing a fully convolutional U-Net based architecture for semantic segmentation of surface cracks which allows for the use of small datasets through a patch based training process. Our proposed configuration makes use of residual connections inside the convolutional blocks as well as including an attention based gating mechanism between the encoder and decoder section of this architecture, which only propagates relevant activations further. Using our proposed architecture we achieve new state of the art results in two different crack datasets, outperforming the previous best results in two metrics each.
Original languageEnglish
Title of host publication2019 IEEE International Conference on Image Processing (ICIP)
PublisherIEEE
Pages1460-1464
Number of pages5
DOIs
Publication statusPublished - 26 Aug 2019

Fingerprint

Pavements
Cracks
Neural networks
Image segmentation
Pixels
Chemical activation
Semantics

Keywords

  • image segmentation
  • training
  • decoding
  • surface cracks
  • semantics
  • convolutional codes
  • task analysis

Cite this

@inproceedings{d73fe653497c49078801258455c1f74e,
title = "A convolutional neural network for pavement surface crack segmentation using residual connections and attention gating",
abstract = "Conventional surface crack segmentation requires images manually labelled by a trained expert. It is a challenging task as cracks can vary in orientation and size, with some parts of cracks only being one pixel wide. Further, available training data for crack segmentation is sparse. In this work we propose to automate this annotation task, by introducing a fully convolutional U-Net based architecture for semantic segmentation of surface cracks which allows for the use of small datasets through a patch based training process. Our proposed configuration makes use of residual connections inside the convolutional blocks as well as including an attention based gating mechanism between the encoder and decoder section of this architecture, which only propagates relevant activations further. Using our proposed architecture we achieve new state of the art results in two different crack datasets, outperforming the previous best results in two metrics each.",
keywords = "image segmentation, training, decoding, surface cracks, semantics, convolutional codes, task analysis",
author = "Jacob Konig and Jenkins, {Mark David} and Peter Barrie and Mike Mannion and Gordon Morison",
note = "Acceptance email req'd 21/11/19 DC",
year = "2019",
month = "8",
day = "26",
doi = "10.1109/ICIP.2019.8803060",
language = "English",
pages = "1460--1464",
booktitle = "2019 IEEE International Conference on Image Processing (ICIP)",
publisher = "IEEE",

}

A convolutional neural network for pavement surface crack segmentation using residual connections and attention gating. / Konig, Jacob; Jenkins, Mark David; Barrie, Peter; Mannion, Mike; Morison, Gordon.

2019 IEEE International Conference on Image Processing (ICIP). IEEE, 2019. p. 1460-1464.

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

TY - GEN

T1 - A convolutional neural network for pavement surface crack segmentation using residual connections and attention gating

AU - Konig, Jacob

AU - Jenkins, Mark David

AU - Barrie, Peter

AU - Mannion, Mike

AU - Morison, Gordon

N1 - Acceptance email req'd 21/11/19 DC

PY - 2019/8/26

Y1 - 2019/8/26

N2 - Conventional surface crack segmentation requires images manually labelled by a trained expert. It is a challenging task as cracks can vary in orientation and size, with some parts of cracks only being one pixel wide. Further, available training data for crack segmentation is sparse. In this work we propose to automate this annotation task, by introducing a fully convolutional U-Net based architecture for semantic segmentation of surface cracks which allows for the use of small datasets through a patch based training process. Our proposed configuration makes use of residual connections inside the convolutional blocks as well as including an attention based gating mechanism between the encoder and decoder section of this architecture, which only propagates relevant activations further. Using our proposed architecture we achieve new state of the art results in two different crack datasets, outperforming the previous best results in two metrics each.

AB - Conventional surface crack segmentation requires images manually labelled by a trained expert. It is a challenging task as cracks can vary in orientation and size, with some parts of cracks only being one pixel wide. Further, available training data for crack segmentation is sparse. In this work we propose to automate this annotation task, by introducing a fully convolutional U-Net based architecture for semantic segmentation of surface cracks which allows for the use of small datasets through a patch based training process. Our proposed configuration makes use of residual connections inside the convolutional blocks as well as including an attention based gating mechanism between the encoder and decoder section of this architecture, which only propagates relevant activations further. Using our proposed architecture we achieve new state of the art results in two different crack datasets, outperforming the previous best results in two metrics each.

KW - image segmentation

KW - training

KW - decoding

KW - surface cracks

KW - semantics

KW - convolutional codes

KW - task analysis

U2 - 10.1109/ICIP.2019.8803060

DO - 10.1109/ICIP.2019.8803060

M3 - Conference contribution

SP - 1460

EP - 1464

BT - 2019 IEEE International Conference on Image Processing (ICIP)

PB - IEEE

ER -