Self-Supervised Learning for Anomaly Detection With Dynamic Local Augmentation

Anomaly detection is an important problem for recent advances in machine learning. To this end, many attempts have emerged to detect unknown anomalies of the images by learning representations and designing score functions. In this paper, we propose a simple yet effective framework for unsupervised...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Seungdong Yoa, Seungjun Lee, Chiyoon Kim, Hyunwoo J Kim
Formato:	article
Lenguaje:	EN
Publicado:	IEEE 2021
Materias:	Anomaly detection computer vision deep learning machine learning self-supervised learning Electrical engineering. Electronics. Nuclear engineering TK1-9971
Acceso en línea:	https://doaj.org/article/c801b504e22a487a82b92f13215e2473
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:c801b504e22a487a82b92f13215e2473
record_format	dspace
spelling	oai:doaj.org-article:c801b504e22a487a82b92f13215e24732021-11-18T00:04:58ZSelf-Supervised Learning for Anomaly Detection With Dynamic Local Augmentation2169-353610.1109/ACCESS.2021.3124525https://doaj.org/article/c801b504e22a487a82b92f13215e24732021-01-01T00:00:00Zhttps://ieeexplore.ieee.org/document/9597511/https://doaj.org/toc/2169-3536Anomaly detection is an important problem for recent advances in machine learning. To this end, many attempts have emerged to detect unknown anomalies of the images by learning representations and designing score functions. In this paper, we propose a simple yet effective framework for unsupervised anomaly detection using self-supervised learning. We extend conventional self-supervised learning for an anomaly detection problem. In anomaly detection, anomalous patterns appear in the local regions of an image, so we employ <italic>dynamic local augmentation</italic> to generate a negative pair of the images from the normal training dataset. Specifically, in addition to learning the global representation of an image, our framework contrasts a normal sample to a locally augmented sample. To effectively apply the local augmentations regardless of a category or a random location of an image, we use dynamically weighted local augmentations to generate more suitable negative samples. We also present a novel scoring function for detecting unseen anomalous patterns. Our experiment demonstrates the effectiveness of our method, and we show that our framework achieves competitive performance compared to state-of-the-art methods on MVTec Anomaly Detection dataset.Seungdong YoaSeungjun LeeChiyoon KimHyunwoo J KimIEEEarticleAnomaly detectioncomputer visiondeep learningmachine learningself-supervised learningElectrical engineering. Electronics. Nuclear engineeringTK1-9971ENIEEE Access, Vol 9, Pp 147201-147211 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	Anomaly detection computer vision deep learning machine learning self-supervised learning Electrical engineering. Electronics. Nuclear engineering TK1-9971
spellingShingle	Anomaly detection computer vision deep learning machine learning self-supervised learning Electrical engineering. Electronics. Nuclear engineering TK1-9971 Seungdong Yoa Seungjun Lee Chiyoon Kim Hyunwoo J Kim Self-Supervised Learning for Anomaly Detection With Dynamic Local Augmentation
description	Anomaly detection is an important problem for recent advances in machine learning. To this end, many attempts have emerged to detect unknown anomalies of the images by learning representations and designing score functions. In this paper, we propose a simple yet effective framework for unsupervised anomaly detection using self-supervised learning. We extend conventional self-supervised learning for an anomaly detection problem. In anomaly detection, anomalous patterns appear in the local regions of an image, so we employ <italic>dynamic local augmentation</italic> to generate a negative pair of the images from the normal training dataset. Specifically, in addition to learning the global representation of an image, our framework contrasts a normal sample to a locally augmented sample. To effectively apply the local augmentations regardless of a category or a random location of an image, we use dynamically weighted local augmentations to generate more suitable negative samples. We also present a novel scoring function for detecting unseen anomalous patterns. Our experiment demonstrates the effectiveness of our method, and we show that our framework achieves competitive performance compared to state-of-the-art methods on MVTec Anomaly Detection dataset.
format	article
author	Seungdong Yoa Seungjun Lee Chiyoon Kim Hyunwoo J Kim
author_facet	Seungdong Yoa Seungjun Lee Chiyoon Kim Hyunwoo J Kim
author_sort	Seungdong Yoa
title	Self-Supervised Learning for Anomaly Detection With Dynamic Local Augmentation
title_short	Self-Supervised Learning for Anomaly Detection With Dynamic Local Augmentation
title_full	Self-Supervised Learning for Anomaly Detection With Dynamic Local Augmentation
title_fullStr	Self-Supervised Learning for Anomaly Detection With Dynamic Local Augmentation
title_full_unstemmed	Self-Supervised Learning for Anomaly Detection With Dynamic Local Augmentation
title_sort	self-supervised learning for anomaly detection with dynamic local augmentation
publisher	IEEE
publishDate	2021
url	https://doaj.org/article/c801b504e22a487a82b92f13215e2473
work_keys_str_mv	AT seungdongyoa selfsupervisedlearningforanomalydetectionwithdynamiclocalaugmentation AT seungjunlee selfsupervisedlearningforanomalydetectionwithdynamiclocalaugmentation AT chiyoonkim selfsupervisedlearningforanomalydetectionwithdynamiclocalaugmentation AT hyunwoojkim selfsupervisedlearningforanomalydetectionwithdynamiclocalaugmentation
_version_	1718425250842017792

Self-Supervised Learning for Anomaly Detection With Dynamic Local Augmentation

Ejemplares similares