Advances in semi and self-supervised learning techniques for cross-domain visual inspection tools
International Journal of Machine Learning and Cybernetics, cilt.17, sa.9, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 17 Sayı: 9
- Basım Tarihi: 2026
- Doi Numarası: 10.1007/s13042-026-03259-2
- Dergi Adı: International Journal of Machine Learning and Cybernetics
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Aerospace Database, Compendex, INSPEC, Materials Science & Engineering Collection (ProQuest), Technology Collection (ProQuest)
- Anahtar Kelimeler: Computer vision, Medical imaging, Object localization, Self-supervised learning, Semi-supervised learning, Visual inspection
- Akdeniz Üniversitesi Adresli: Evet
Özet
Semi-supervised and self-supervised learning (SSL and Self-SL) have gained increasing attention in visual inspection tasks such as anomaly detection, defect identification, and threat localization. Across domains including medical imaging, industrial inspection, and baggage screening, these approaches offer the potential to reduce dependence on large annotated datasets while achieving performance comparable to fully supervised models. This review critically synthesizes recent advancements in label-efficient learning paradigms and examines how these techniques address challenges such as occlusion, clutter, limited annotations, and domain variability. The work is a rigorous systematic review of over 170 peer-reviewed articles published no earlier than 2016, retrieved in one of the largest scholarly databases such as IEEE Xplore, ACM Digital library, Springer1ink, and ScienceDirect. The screening process was done using a predetermined set of inclusion and exclusion criteria, which helped to make it methodologically objective and transparent. The reason behind this review is the high cost, time, and professional verification of annotating real-life inspection datasets, especially in such high-stakes sectors as medical testing, plant equipment fault detection, and airport security. Through an analysis of the advances in quality, strength, computing efficiency, reduction of annotations, and cross-domain transfer of improved results, this review identifies the current research that has been made possible through recent advances in SSL and Self-SL methods, which support real-world visual inspection processes. The review also discovers cohesive tendencies across fields, limits that cannot be overcome such as pseudo-label noise, domain shift, safety threats, and interpretability issues, and future prospects to do research like multimodal fusion, which uses transformers to perform self-supervision, and massively radially-unified cross-domain adaptation.