Details

Analysis of continual unsupervised learning for surface anomaly detection
ID Ivanovska, Sara (Author), ID Skočaj, Danijel (Mentor) More about this mentor... This link opens in a new window

.pdfPDF - Presentation file, Download (8,01 MB)
MD5: F1543A93864A729EEB983623BAFE3D4E

Abstract
Industrial anomaly detection systems are typically built using separate models for each product category, resulting in substantial computational demands and ongoing maintenance challenges. This thesis explores continual learning as an alternative strategy, enabling a single model to sequentially learn and retain knowledge across multiple product categories. We conduct a comprehensive evaluation of three state-of-the-art methods— PatchCore, EfficientAD, and SuperSimpleNet—under three training paradigms: separate models, joint training, and continual learning, using the MVTec AD dataset. The results show that continual learning can maintain robust performance, with PatchCore maintaining strong performance in continual learning compared to separate models. Moreover, memory-based approaches consistently outperform replay buffer strategies. These insights offer practical guidance for deploying anomaly detection systems in dynamic industrial environments, highlighting continual learning as a compelling and resource-efficient alternative to the traditional use of multiple dedicated models.

Language:English
Keywords:continual learning, anomaly detection, industrial inspection, memory bank, replay buffer, MVTec AD
Work type:Bachelor thesis/paper
Typology:2.11 - Undergraduate Thesis
Organization:FRI - Faculty of Computer and Information Science
Year:2025
PID:20.500.12556/RUL-171948 This link opens in a new window
COBISS.SI-ID:248574211 This link opens in a new window
Publication date in RUL:04.09.2025
Views:920
Downloads:440
Metadata:XML DC-XML DC-RDF
:
Copy citation
Share:Bookmark and Share

Secondary language

Language:Slovenian
Title:Analiza sprotnega nenadzorovanega učenja za detekcijo površinskih anomalij
Abstract:
Sistemi za detekcijo industrijskih anomalij se običajno gradijo z uporabo ločenih modelov za vsako kategorijo izdelkov, kar povzroča znatne računske zahteve in težave z vzdrževanjem. Ta diplomska naloga raziskuje sprotno učenje kot alternativno strategijo, ki omogoča enemu modelu, da se zaporedno uči in ohranja znanje o več kategorijah izdelkov. Izvedli smo celovito ovrednotenje treh sodobnih metod—PatchCore, EfficientAD in SuperSimpleNet—pod tremi paradigmami učenja: ločeni modeli, skupno učenje in sprotno učenje, z uporabo podatkovne množice MVTec AD. Rezultati kažejo, da lahko sprotno učenje ohrani robustno delovanje, pri čemer PatchCore ohranja močno uspešnost pri sprotnem učenju v primerjavi z ločenimi modeli. Poleg tega pristopi na osnovi spomina dosledno presegajo strategije ponovnega predvajanja. Te ugotovitve ponujajo praktične smernice za uvajanje sistemov za detekcijo anomalij v dinamičnih industrijskih okoljih in poudarjajo sprotno učenje kot prepričljivo in učinkovito alternativo tradicionalni uporabi več-namenskih modelov.

Keywords:sprotno učenje, detekcija anomalij, industrijski pregled, spominska banka, medpomnilnik za ponovno predvajanje, MVTec AD

Similar documents

Similar works from RUL:
Similar works from other Slovenian collections:

Back