File Exchange

image thumbnail

Crack detection using one-class SVM (1クラスsvmによるひび割れ検知)

version 1.0.2 (3.06 MB) by Kenta
This demo shows how to detect the crack images using one-class SVM. このデモでは、1クラスSVMを用いて、ひび割れを自動的に検知します。


Updated 11 May 2020

View Version History

View License

This demo shows how to detect the crack images using one-class SVM. In anomaly detection, normal images can be obtained a lot while the anomaly images are not frequenctly obtained; we cannot get sufficient number of training image of the anomaly data. In that case, a classifier was trained only with normal images and the anomaly images are detected when the pattern is different from the one it has learnt.
If sufficient number of anomaly images are available, crack/normal images can be discriminated using deep learning. For the demo, please check this file (
In this demo, we use a dataset of concrete crack images introduced by L Zhang [1]. The data is available at [2].
A portion of this code was obtained from Deep Learning Evaluatio Kit located here [3].

このデモでは、1クラスSVM(サポートベクターマシン)を用いて、コンクリートのひび割れを自動的に検知します。合計4万枚の通常/ひび割れ画像を含むデータセット [1] を用いて、訓練や検証を行います。CNN(畳み込みニューラルネットワーク)などとは異なり、数分で学習を完了することができます。

[Key words]
anomaly detection, crack, detection, fissure, less data, support vector machine (SVM), unsupervised learning, 1-class SVM

[1] Zhang, Lei, et al. "Road crack detection using deep convolutional neural network." 2016 IEEE international conference on image processing (ICIP). IEEE, 2016.
[2] Concrete Crack Images for Classification (
[3] Takuji Fukumoto (2020). ディープラーニング評価キット [画像分類用] (, MATLAB Central File Exchange. Retrieved May 5, 2020.

Cite As

Kenta (2021). Crack detection using one-class SVM (1クラスsvmによるひび割れ検知) (, MATLAB Central File Exchange. Retrieved .

Comments and Ratings (3)

omobayo esan

can you please express this code in English

Tohru Kikawada


MATLAB Release Compatibility
Created with R2020a
Compatible with any release
Platform Compatibility
Windows macOS Linux

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!