{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,25]],"date-time":"2026-03-25T06:05:46Z","timestamp":1774418746502,"version":"3.50.1"},"reference-count":26,"publisher":"MDPI AG","issue":"6","license":[{"start":{"date-parts":[[2026,3,20]],"date-time":"2026-03-20T00:00:00Z","timestamp":1773964800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Applied Sciences"],"abstract":"<jats:p>Accurate surface defect classification is critical for industrial quality control. Although Deep Learning achieves strong results on individual datasets, most prior studies benchmark only a narrow set of models under inconsistent pipelines, limiting comparability and industrial relevance. This work introduces the first systematic benchmark of ten architectures\u2014CNNs (CNN, ResNet18\/50), lightweight models (MobileNetV2, SuperSimpleNet, GhostNet, EfficientNetV2), Vision Transformers (Swin Transformer), a hybrid CNN\u2013Transformer (CoAtNet), and a one-stage detector (YOLOv12)\u2014across five public defect datasets (NEU-DET, X-SDD, KolektorSDD2, DAGM, MTDD) under a unified pipeline. Results show that Swin Transformer and CoAtNet achieve the best performance (mean F1-scores 90.8% and 85.5%), while EfficientNetV2 underperformed (41.9%), underscoring the need for domain-specific benchmarks. Lightweight models such as MobileNetV2, GhostNet, and SuperSimpleNet deliver competitive accuracy at much lower cost, offering practical solutions for edge deployment. By bridging the gap between academic benchmarks and manufacturing requirements, this study provides actionable guidance for selecting defect detection models in automated inspection.<\/jats:p>","DOI":"10.3390\/app16063022","type":"journal-article","created":{"date-parts":[[2026,3,20]],"date-time":"2026-03-20T17:02:43Z","timestamp":1774026163000},"page":"3022","update-policy":"https:\/\/doi.org\/10.3390\/mdpi_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Cross-Dataset Benchmarking of Deep Learning Models for Surface Defect Classification in Metal Parts"],"prefix":"10.3390","volume":"16","author":[{"ORCID":"https:\/\/orcid.org\/0009-0000-9858-2018","authenticated-orcid":false,"given":"F\u00e1bio Mendes da","family":"Silva","sequence":"first","affiliation":[{"name":"Faculdade de Engenharia, Universidade do Porto (FEUP), Rua Dr. Roberto Frias, 4200-465 Porto, Portugal"},{"name":"INEGI\u2014Institute of Science and Innovation in Mechanical and Industrial Engineering, Rua Dr. Roberto Frias, 4200-465 Porto, Portugal"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7603-6526","authenticated-orcid":false,"given":"Jo\u00e3o Manuel R. S.","family":"Tavares","sequence":"additional","affiliation":[{"name":"Faculdade de Engenharia, Universidade do Porto (FEUP), Rua Dr. Roberto Frias, 4200-465 Porto, Portugal"},{"name":"INEGI\u2014Institute of Science and Innovation in Mechanical and Industrial Engineering, Rua Dr. Roberto Frias, 4200-465 Porto, Portugal"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7359-4370","authenticated-orcid":false,"given":"Ant\u00f3nio Mendes","family":"Lopes","sequence":"additional","affiliation":[{"name":"Faculdade de Engenharia, Universidade do Porto (FEUP), Rua Dr. Roberto Frias, 4200-465 Porto, Portugal"},{"name":"INEGI\u2014Institute of Science and Innovation in Mechanical and Industrial Engineering, Rua Dr. Roberto Frias, 4200-465 Porto, Portugal"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4146-6224","authenticated-orcid":false,"given":"Antonio Ramos","family":"Silva","sequence":"additional","affiliation":[{"name":"Faculdade de Engenharia, Universidade do Porto (FEUP), Rua Dr. Roberto Frias, 4200-465 Porto, Portugal"},{"name":"INEGI\u2014Institute of Science and Innovation in Mechanical and Industrial Engineering, Rua Dr. Roberto Frias, 4200-465 Porto, Portugal"}]}],"member":"1968","published-online":{"date-parts":[[2026,3,20]]},"reference":[{"key":"ref_1","doi-asserted-by":"crossref","unstructured":"Feng, X., Gao, X., and Luo, L. (2021). A ResNet50-Based Method for Classifying Surface Defects in Hot-Rolled Strip Steel. Mathematics, 9.","DOI":"10.3390\/math9192359"},{"key":"ref_2","doi-asserted-by":"crossref","first-page":"31941","DOI":"10.1038\/s41598-024-83430-3","article-title":"Feature optimization-guided high-precision and real-time metal surface defect detection network","volume":"14","author":"Chan","year":"2024","journal-title":"Sci. Rep."},{"key":"ref_3","doi-asserted-by":"crossref","first-page":"8966","DOI":"10.1038\/s41598-025-93469-5","article-title":"A lightweight algorithm for steel surface defect detection using improved YOLOv8","volume":"15","author":"Ma","year":"2025","journal-title":"Sci. Rep."},{"key":"ref_4","doi-asserted-by":"crossref","unstructured":"Konovalenko, I., Maruschak, P., Brezinov\u00e1, J., Prentkovskis, O., and Brezina, J. (2022). Research of U-Net-Based CNN Architectures for Metal Surface Defect Detection. Machines, 10.","DOI":"10.3390\/machines10050327"},{"key":"ref_5","doi-asserted-by":"crossref","unstructured":"Liu, Y., Qin, Y., Lin, Z., Xia, H., and Wang, C. (2024). Detection of Scratch Defects on Metal Surfaces Based on MSDD-UNet. Electronics, 13.","DOI":"10.3390\/electronics13163241"},{"key":"ref_6","doi-asserted-by":"crossref","unstructured":"Shiferaw, T.G., and Yao, L. (2024). Autoencoder-Based Unsupervised Surface Defect Detection Using Two-Stage Training. J. Imaging, 10.","DOI":"10.3390\/jimaging10050111"},{"key":"ref_7","unstructured":"Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A.N., Kaiser, \u0141., and Polosukhin, I. (2023). Attention Is All You Need. arXiv."},{"key":"ref_8","doi-asserted-by":"crossref","first-page":"155","DOI":"10.1007\/s11227-024-06662-0","article-title":"RFAConv-CBM-ViT: Enhanced vision transformer for metal surface defect detection","volume":"81","author":"Wei","year":"2024","journal-title":"J. Supercomput."},{"key":"ref_9","doi-asserted-by":"crossref","first-page":"11105","DOI":"10.1038\/s41598-025-94936-9","article-title":"Metal surface defect detection using SLF-YOLO enhanced YOLOv8 model","volume":"15","author":"Liu","year":"2025","journal-title":"Sci. Rep."},{"key":"ref_10","doi-asserted-by":"crossref","unstructured":"Wi, T., Yang, M., Park, S., and Jeong, J. (2024). D2-SPDM: Faster R-CNN-Based Defect Detection and Surface Pixel Defect Mapping with Label Enhancement in Steel Manufacturing Processes. Appl. Sci., 14.","DOI":"10.3390\/app14219836"},{"key":"ref_11","doi-asserted-by":"crossref","first-page":"108482","DOI":"10.1016\/j.engappai.2024.108482","article-title":"LAACNet: Lightweight adaptive activation convolution network-based defect detection on polished metal surfaces","volume":"133","author":"Lv","year":"2024","journal-title":"Eng. Appl. Artif. Intell."},{"key":"ref_12","doi-asserted-by":"crossref","first-page":"112754","DOI":"10.1016\/j.engappai.2025.112754","article-title":"Multi-scale target detection of metal surface defects in additive manufacturing based on reinforcement learning","volume":"162","author":"Niu","year":"2025","journal-title":"Eng. Appl. Artif. Intell."},{"key":"ref_13","unstructured":"Northeastern University (2025, July 09). NEU-DET Surface Defect Database, 2024. Available online: http:\/\/faculty.neu.edu.cn\/yunhyan\/NEU_surface_defect_database.html."},{"key":"ref_14","unstructured":"Aligah, S.A. (2025, July 09). X-SDD Dataset, Kaggle, 2023. Available online: https:\/\/www.kaggle.com\/datasets\/sayelabualigah\/x-sdd."},{"key":"ref_15","doi-asserted-by":"crossref","first-page":"103459","DOI":"10.1016\/j.compind.2021.103459","article-title":"Mixed supervision for surface-defect detection: From weakly to fully supervised learning","volume":"129","author":"Tabernik","year":"2021","journal-title":"Comput. Ind."},{"key":"ref_16","unstructured":"Wieler, M., Hahn, T., and Hamprecht, F.A. (2007). Weakly Supervised Learning for Industrial Optical Inspection, Collaboratory for Image Processing (HCI)."},{"key":"ref_17","doi-asserted-by":"crossref","unstructured":"Carvalho, P., Durupt, A., and Grandvalet, Y. (2022). A Survey of Machine Learning Approaches for Visual Inspection on the DAGM Dataset. Advances in Manufacturing Technology XXXV, IOP.","DOI":"10.3233\/ATDE220600"},{"key":"ref_18","unstructured":"Huang, Y., Qiu, C., Guo, Y., Wang, X., and Yuan, K. (2024). Surface Defect Saliency of Magnetic Tile. IEEE International Conference on Automation Science and Engineering (CASE), IEEE."},{"key":"ref_19","doi-asserted-by":"crossref","unstructured":"He, K., Zhang, X., Ren, S., and Sun, J. (2016). Deep Residual Learning for Image Recognition. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), IEEE.","DOI":"10.1109\/CVPR.2016.90"},{"key":"ref_20","doi-asserted-by":"crossref","unstructured":"Sandler, M., Howard, A.G., Zhu, M., Zhmoginov, A., and Chen, L.-C. (2018). MobileNetV2: Inverted Residuals and Linear Bottlenecks. IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR), IEEE.","DOI":"10.1109\/CVPR.2018.00474"},{"key":"ref_21","doi-asserted-by":"crossref","unstructured":"Rolih, B., Fu\u010dka, M., and Sko\u010daj, D. (2024). SuperSimpleNet: Unifying Unsupervised and Supervised Learning for Fast and Reliable Surface Defect Detection. International Conference on Pattern Recognition, Springer.","DOI":"10.1007\/978-3-031-78192-6_4"},{"key":"ref_22","doi-asserted-by":"crossref","unstructured":"Liu, Z., Lin, Y., Cao, Y., Hu, H., Wei, Y., Zhang, Z., Lin, S., and Guo, B. (2021). Swin Transformer: Hierarchical Vision Transformer using Shifted Windows. Proceedings of the IEEE\/CVF International Conference on Computer Vision, IEEE.","DOI":"10.1109\/ICCV48922.2021.00986"},{"key":"ref_23","unstructured":"Tan, M., and Le, Q.V. (2021). EfficientNetV2: Smaller Models and Faster Training. arXiv."},{"key":"ref_24","doi-asserted-by":"crossref","unstructured":"Han, K., Wang, Y., Tian, Q., Guo, J., Xu, C., and Xu, C. (2020). GhostNet: More Features from Cheap Operations. arXiv.","DOI":"10.1109\/CVPR42600.2020.00165"},{"key":"ref_25","unstructured":"Dai, Z., Liu, H., Le, Q.V., and Tan, M. (2021). CoAtNet: Marrying Convolution and Attention for All Data Sizes. arXiv."},{"key":"ref_26","unstructured":"Tian, Y., Ye, Q., and Doermann, D. (2025). YOLOv12: Attention-Centric Real-Time Object Detectors. arXiv."}],"container-title":["Applied Sciences"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.mdpi.com\/2076-3417\/16\/6\/3022\/pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,25]],"date-time":"2026-03-25T05:14:19Z","timestamp":1774415659000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.mdpi.com\/2076-3417\/16\/6\/3022"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,3,20]]},"references-count":26,"journal-issue":{"issue":"6","published-online":{"date-parts":[[2026,3]]}},"alternative-id":["app16063022"],"URL":"https:\/\/doi.org\/10.3390\/app16063022","relation":{},"ISSN":["2076-3417"],"issn-type":[{"value":"2076-3417","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,3,20]]}}}