Fractals as Pre-Training Datasets for Anomaly Detection and Localization

Anomaly detection is crucial in large-scale industrial manufacturing as it helps to detect and localize defective parts. Pre-training feature extractors on large-scale datasets is a popular approach for this task. Stringent data security, privacy regulations, high costs, and long acquisition time hi...

Full description

Saved in:
Bibliographic Details
Main Authors: Cynthia I. Ugwu, Emanuele Caruso, Oswald Lanz
Format: Article
Language:English
Published: MDPI AG 2024-11-01
Series:Fractal and Fractional
Subjects:
Online Access:https://www.mdpi.com/2504-3110/8/11/661
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846153540717772800
author Cynthia I. Ugwu
Emanuele Caruso
Oswald Lanz
author_facet Cynthia I. Ugwu
Emanuele Caruso
Oswald Lanz
author_sort Cynthia I. Ugwu
collection DOAJ
description Anomaly detection is crucial in large-scale industrial manufacturing as it helps to detect and localize defective parts. Pre-training feature extractors on large-scale datasets is a popular approach for this task. Stringent data security, privacy regulations, high costs, and long acquisition time hinder the development of large-scale datasets for training and benchmarking. Despite recent work focusing primarily on the development of new anomaly detection methods based on such extractors, not much attention has been paid to the importance of the data used for pre-training. This study compares representative models pre-trained with fractal images against those pre-trained with ImageNet, without subsequent task-specific fine-tuning. We evaluated the performance of eleven state-of-the-art methods on MVTecAD, MVTec LOCO AD, and VisA, well-known benchmark datasets inspired by real-world industrial inspection scenarios. Further, we propose a novel method to create a dataset by combining the dynamically generated fractal images creating a “Multi-Formula” dataset. Even though pre-training with ImageNet leads to better results, fractals can achieve close performance to ImageNet under proper parametrization. This opens up the possibility for a new research direction where feature extractors could be trained on synthetically generated abstract datasets mitigating the ever-increasing demand for data in machine learning while circumventing privacy and security concerns.
format Article
id doaj-art-cb78bf7448f046a8a301971787df4d2c
institution Kabale University
issn 2504-3110
language English
publishDate 2024-11-01
publisher MDPI AG
record_format Article
series Fractal and Fractional
spelling doaj-art-cb78bf7448f046a8a301971787df4d2c2024-11-26T18:05:04ZengMDPI AGFractal and Fractional2504-31102024-11-0181166110.3390/fractalfract8110661Fractals as Pre-Training Datasets for Anomaly Detection and LocalizationCynthia I. Ugwu0Emanuele Caruso1Oswald Lanz2Department of Engineering, Free University of Bozen-Bolzano, 39100 Bozen-Bolzano, ItalyDepartment of Engineering, Free University of Bozen-Bolzano, 39100 Bozen-Bolzano, ItalyDepartment of Engineering, Free University of Bozen-Bolzano, 39100 Bozen-Bolzano, ItalyAnomaly detection is crucial in large-scale industrial manufacturing as it helps to detect and localize defective parts. Pre-training feature extractors on large-scale datasets is a popular approach for this task. Stringent data security, privacy regulations, high costs, and long acquisition time hinder the development of large-scale datasets for training and benchmarking. Despite recent work focusing primarily on the development of new anomaly detection methods based on such extractors, not much attention has been paid to the importance of the data used for pre-training. This study compares representative models pre-trained with fractal images against those pre-trained with ImageNet, without subsequent task-specific fine-tuning. We evaluated the performance of eleven state-of-the-art methods on MVTecAD, MVTec LOCO AD, and VisA, well-known benchmark datasets inspired by real-world industrial inspection scenarios. Further, we propose a novel method to create a dataset by combining the dynamically generated fractal images creating a “Multi-Formula” dataset. Even though pre-training with ImageNet leads to better results, fractals can achieve close performance to ImageNet under proper parametrization. This opens up the possibility for a new research direction where feature extractors could be trained on synthetically generated abstract datasets mitigating the ever-increasing demand for data in machine learning while circumventing privacy and security concerns.https://www.mdpi.com/2504-3110/8/11/661fractalsMandelbulbanomaly detectionindustrial inspectionsynthetic data
spellingShingle Cynthia I. Ugwu
Emanuele Caruso
Oswald Lanz
Fractals as Pre-Training Datasets for Anomaly Detection and Localization
Fractal and Fractional
fractals
Mandelbulb
anomaly detection
industrial inspection
synthetic data
title Fractals as Pre-Training Datasets for Anomaly Detection and Localization
title_full Fractals as Pre-Training Datasets for Anomaly Detection and Localization
title_fullStr Fractals as Pre-Training Datasets for Anomaly Detection and Localization
title_full_unstemmed Fractals as Pre-Training Datasets for Anomaly Detection and Localization
title_short Fractals as Pre-Training Datasets for Anomaly Detection and Localization
title_sort fractals as pre training datasets for anomaly detection and localization
topic fractals
Mandelbulb
anomaly detection
industrial inspection
synthetic data
url https://www.mdpi.com/2504-3110/8/11/661
work_keys_str_mv AT cynthiaiugwu fractalsaspretrainingdatasetsforanomalydetectionandlocalization
AT emanuelecaruso fractalsaspretrainingdatasetsforanomalydetectionandlocalization
AT oswaldlanz fractalsaspretrainingdatasetsforanomalydetectionandlocalization