Domain Transfer for Surface Defect Detection using Few-Shot Learning on Scarce Data
Autor(en): | Gerschner, Felix Paul, Jonas Schmid, Lukas Barthel, Nico Gouromichos, Victor Schmid, Florian Atzmueller, Martin Theissler, Andreas |
Herausgeber: | Dorksen, H. Scanzio, S. Jasperneite, J. Wisniewski, L. Man, K.F. Sauter, T. Seno, L. Trsek, H. Vyatkin, V. |
Stichwörter: | Data scarcity; Data set; Defects inspections; explainable AI; few-shot learning; industrial defect inspection; Learning models; Learning systems; Production efficiency; scarce data; surface defect detection; Surface defect detections; Surface defects; transfer learning | Erscheinungsdatum: | 2023 | Herausgeber: | Institute of Electrical and Electronics Engineers Inc. | Journal: | IEEE International Conference on Industrial Informatics (INDIN) | Volumen: | 2023-July | Zusammenfassung: | This study evaluates the effectiveness of transfer learning models in industrial surface defect detection using few-shot learning. Surface defect detection is a critical task in various industrial applications, where accurately detecting and classifying defects can improve product quality and increase manufacturing efficiency. However, data scarcity is a considerable challenge: obtaining and labelling defect samples is a costly, time-consuming process and difficult due to their infrequent occurrence. Few-Shot learning aims to effectively train models using only a limited number of labelled samples, thus mitigating the impact of data scarcity. This study compares the performance of transfer learning models pre-trained on three different data sets for few-shot learning in the context of surface defect detection. On the one hand, transfer learning models pre-trained on the ImageNet data set yield the best overall results in terms of accuracy. On the other hand, our results indicate that the DAGM data set, an industrial optical inspection data set which is close to the target domain, is particularly effective for training models to clearly detect surface defects in a few-shot learning scenario. © 2023 IEEE. |
Beschreibung: | Cited by: 0; Conference name: 21st IEEE International Conference on Industrial Informatics, INDIN 2023; Conference date: 17 July 2023 through 20 July 2023; Conference code: 192026 |
ISBN: | 9781665493130 | ISSN: | 1935-4576 | DOI: | 10.1109/INDIN51400.2023.10217859 | Externe URL: | https://www.scopus.com/inward/record.uri?eid=2-s2.0-85171196774&doi=10.1109%2fINDIN51400.2023.10217859&partnerID=40&md5=56f7811b5636f6f6edfaf1b0881ee1ec |
Zur Langanzeige
Seitenaufrufe
1
Letzte Woche
0
0
Letzter Monat
1
1
geprüft am 17.05.2024