{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,13]],"date-time":"2026-07-13T09:04:39Z","timestamp":1783933479948,"version":"3.55.0"},"reference-count":39,"publisher":"MDPI AG","issue":"13","license":[{"start":{"date-parts":[[2026,7,4]],"date-time":"2026-07-04T00:00:00Z","timestamp":1783123200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Sensors"],"abstract":"<jats:p>This study proposes a deep learning-based framework for the automated classification of photovoltaic solar cells as defective or normal using electroluminescence (EL) imaging. A balanced dataset containing 20,400 EL images, comprising 10,200 defective and 10,200 normal solar cells, was used for model development and evaluation. To reflect practical inspection requirements, cracked and broken cells were combined into a single defective category, resulting in a binary classification task. The dataset includes both monocrystalline and polycrystalline solar cells, which were analyzed together within a unified classification framework to improve applicability to real-world photovoltaic systems. To ensure a fair and unbiased evaluation, dataset partitioning was performed prior to any preprocessing or augmentation operations, and each image was assigned exclusively to the training, validation, or test subset. Data augmentation was applied only to the training set, eliminating the possibility of data leakage. Four state-of-the-art deep learning architectures, EfficientNet-B2, ConvNeXt-Tiny, MaxViT-T, and ResNet-50, were trained and evaluated under identical experimental conditions using the same preprocessing pipeline, training strategy, and dataset split. Model performance was assessed using accuracy, precision, recall, F1-score, ROC-AUC, confusion matrices, and explainability-based activation and attention heat maps. All evaluated models achieved classification accuracies exceeding 98%, demonstrating strong capability for EL-based defect detection. EfficientNet-B2 achieved the highest numerical performance, reaching 99.31% accuracy, 0.9931 F1-score, and 0.9987 ROC-AUC. MaxViT-T exhibited similarly strong performance with rapid convergence and balanced class-wise metrics, while ConvNeXt-Tiny and ResNet-50 also produced highly reliable results. Heat map visualizations revealed that EfficientNet-B2 and MaxViT-T concentrated their attention more precisely on defect regions such as cracks and fractures, providing visual interpretability in addition to quantitative performance. The results demonstrate that modern deep learning architectures can accurately and reliably detect photovoltaic cell defects from EL images under a unified binary classification framework. Furthermore, explainability techniques enhance the transparency of model predictions, supporting the practical deployment of intelligent inspection systems for photovoltaic manufacturing and maintenance applications.<\/jats:p>","DOI":"10.3390\/s26134256","type":"journal-article","created":{"date-parts":[[2026,7,6]],"date-time":"2026-07-06T12:01:37Z","timestamp":1783339297000},"page":"4256","update-policy":"https:\/\/doi.org\/10.3390\/mdpi_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Benchmarking Modern Deep Learning Models for Electroluminescence-Based Solar Cell Defect Detection"],"prefix":"10.3390","volume":"26","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-2862-2453","authenticated-orcid":false,"given":"G\u00f6khan","family":"\u015eahin","sequence":"first","affiliation":[{"name":"Copernicus Institute of Sustainable Development, Utrecht University, Princetonlaan 8A, 3584 CB Utrecht, The Netherlands"},{"name":"Municipality of Dronten, De Rede, 1, 8251 ER Dronten, The Netherlands"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0008-2247-5898","authenticated-orcid":false,"given":"Ali Cengiz","family":"R\u00fcstemli","sequence":"additional","affiliation":[{"name":"Software Engineering Department, Engineering Faculty, Ostim Teknik University, 06000 Ankara, T\u00fcrkiye"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ahmed Yaseen Bishree","family":"Al-Ani","sequence":"additional","affiliation":[{"name":"Electrical-Electronics Engineering Department, Engineering and Architecture Faculty, Bitlis Eren University, 13000 Bitlis, Turkey"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Sabir","family":"R\u00fcstemli","sequence":"additional","affiliation":[{"name":"Electrical-Electronics Engineering Department, Engineering and Architecture Faculty, Bitlis Eren University, 13000 Bitlis, Turkey"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Erdal","family":"Akin","sequence":"additional","affiliation":[{"name":"Department of Computer Science and Media Technology, Malm\u00f6 University, 205 06 Malm\u00f6, Sweden"},{"name":"Department of Computer Engineering, Bitlis Eren University, 13100 Bitlis, T\u00fcrkiye"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1968","published-online":{"date-parts":[[2026,7,4]]},"reference":[{"key":"ref_1","doi-asserted-by":"crossref","first-page":"114810","DOI":"10.1016\/j.eswa.2021.114810","article-title":"Efficient deep feature extraction and classification for identifying defective photovoltaic module cells in Electroluminescence images","volume":"175","author":"Demirci","year":"2021","journal-title":"Expert Syst. Appl."},{"key":"ref_2","unstructured":"Buerhop-Lutz, C., Deitsch, S., Maier, A., Gallwitz, F., Berger, S., Doll, B., and Brabec, C.J. (2018, January 24\u201328). A benchmark for visual identification of defective solar cells in electroluminescence imagery. Proceedings of the 35th European PV Solar Energy Conference and Exhibition, Brussels, Belgium."},{"key":"ref_3","doi-asserted-by":"crossref","first-page":"455","DOI":"10.1016\/j.solener.2019.02.067","article-title":"Automatic classification of defective photovoltaic module cells in electroluminescence images","volume":"185","author":"Deitsch","year":"2019","journal-title":"Sol. Energy"},{"key":"ref_4","doi-asserted-by":"crossref","first-page":"22","DOI":"10.1016\/j.optlaseng.2019.01.016","article-title":"Accurate and robust crack detection using steerable evidence filtering in electroluminescence images of solar cells","volume":"118","author":"Chen","year":"2019","journal-title":"Opt. Lasers Eng."},{"key":"ref_5","doi-asserted-by":"crossref","first-page":"643","DOI":"10.1016\/j.solener.2020.08.027","article-title":"A machine learning framework to identify the hotspot in photovoltaic module using infrared thermography","volume":"208","author":"Ali","year":"2020","journal-title":"Sol. Energy"},{"key":"ref_6","doi-asserted-by":"crossref","first-page":"1211","DOI":"10.1016\/j.renene.2021.06.086","article-title":"Defect detection and quantification in electroluminescence images of solar PV modules using U-net semantic segmentation","volume":"178","author":"Pratt","year":"2021","journal-title":"Renew. Energy"},{"key":"ref_7","first-page":"110786","article-title":"Machine vision based fault diagnosis of photovoltaic modules using lazy learning approach","volume":"191","author":"Sugumaran","year":"2022","journal-title":"Meas. J. Int. Meas. Confed."},{"key":"ref_8","doi-asserted-by":"crossref","unstructured":"Tu, Z., Talebi, H., Zhang, H., Yang, F., Milanfar, P., Bovik, A., and Li, Y. (2022). MaxViT: Multi-Axis Vision Transformer. arXiv.","DOI":"10.1007\/978-3-031-20053-3_27"},{"key":"ref_9","doi-asserted-by":"crossref","unstructured":"Liu, Z., Mao, H., Wu, C.-Y., Feichtenhofer, C., Darrell, T., and Xie, S. (2022). ConvNeXt: A ConvNet for the 2020s. arXiv.","DOI":"10.1109\/CVPR52688.2022.01167"},{"key":"ref_10","unstructured":"Tan, M., and Le, Q.V. (2019, January 9\u201315). EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks. Proceedings of the 36th International Conference on Machine Learning (ICML), Long Beach, CA, USA."},{"key":"ref_11","unstructured":"He, K., Zhang, X., Ren, S., and Sun, J. (2015, January 7\u201312). Deep Residual Learning for Image Recognition. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Boston, MA, USA."},{"key":"ref_12","doi-asserted-by":"crossref","first-page":"132","DOI":"10.1016\/j.neunet.2020.04.012","article-title":"Micro-cracks detection of solar cells surface via combining short-term and long-term deep features","volume":"127","author":"Qian","year":"2020","journal-title":"Neural Netw."},{"key":"ref_13","doi-asserted-by":"crossref","unstructured":"Mathias, N., Shaikh, F., Thakur, C., Shetty, S., Dumane, P., and Chavan, D. (2020, January 8\u201310). Detection of Micro-Cracks in Electroluminescence Images of Photovoltaic Modules. Proceedings of the 3rd International Conference on Advances in Science & Technology (ICAST), Bahir Dar, Ethiopia.","DOI":"10.2139\/ssrn.3563821"},{"key":"ref_14","doi-asserted-by":"crossref","unstructured":"Wang, J., Bi, L., Sun, P., Jiao, X., Ma, X., Lei, X., and Luo, Y. (2023). Deep-learning-based automatic detection of photovoltaic cell defects in electroluminescence images. Sensors, 23.","DOI":"10.3390\/s23010297"},{"key":"ref_15","unstructured":"Bothe, K., Pohl, P., Schmidt, J., Weber, T., Altermatt, P., Fischer, B., and Brendel, R. (2006, January 4\u20138). Electroluminescence Imaging as an In-Line Characterisation Tool for Solar Cell Production. Proceedings of the 21st European Photovoltaik Solar Energy Conference, Dresden, Germany."},{"key":"ref_16","doi-asserted-by":"crossref","first-page":"100442","DOI":"10.1016\/j.array.2025.100442","article-title":"Dual CNN for photovoltaic electroluminescence images microcrack detection","volume":"27","author":"Samrouth","year":"2025","journal-title":"Array"},{"key":"ref_17","doi-asserted-by":"crossref","unstructured":"Amiri, A.F., Kichou, S., Oudira, H., Chouder, A., and Silvestre, S. (2024). Fault Detection and Diagnosis of a Photovoltaic System Based on Deep Learning Using the Combination of a Convolutional Neural Network (CNN) and Bidirectional Gated Recurrent Unit (Bi-GRU). Sustainability, 16.","DOI":"10.3390\/su16031012"},{"key":"ref_18","doi-asserted-by":"crossref","first-page":"471","DOI":"10.1002\/pip.3395","article-title":"Deep Learning-Based Automatic Detection of Multitype Defects in Photovoltaic Modules and Application in Real Production Line","volume":"29","author":"Zhao","year":"2021","journal-title":"Prog. Photovolt. Res. Appl."},{"key":"ref_19","doi-asserted-by":"crossref","unstructured":"Akta\u015f, M., Do\u011fan, F., and T\u00fcrko\u011flu, \u0130. (2026). A Hybrid CNN\u2013Transformer Approach for Photovoltaic Cell Defect Classification Using Electroluminescence Imaging. Sensors, 26.","DOI":"10.3390\/s26082450"},{"key":"ref_20","doi-asserted-by":"crossref","first-page":"118866","DOI":"10.1016\/j.enconman.2024.118866","article-title":"Photovoltaic Modules Fault Detection, Power Output, and Parameter Estimation: A Deep Learning Approach Based on Electroluminescence Images","volume":"319","author":"Ozturk","year":"2024","journal-title":"Energy Convers. Manag."},{"key":"ref_21","doi-asserted-by":"crossref","first-page":"102001","DOI":"10.1088\/1361-6501\/ad6203","article-title":"A survey on fault diagnosis of rotating machinery based on machine learning","volume":"35","author":"Wang","year":"2024","journal-title":"Meas. Sci. Technol."},{"key":"ref_22","doi-asserted-by":"crossref","first-page":"110019","DOI":"10.1016\/j.jobe.2024.110019","article-title":"Enhancing photovoltaic solar panel performance with integration of PCM-based spectral filter and self-cleaning coating","volume":"94","author":"Sheikholeslami","year":"2024","journal-title":"J. Build. Eng."},{"key":"ref_23","doi-asserted-by":"crossref","first-page":"123851","DOI":"10.1016\/j.applthermaleng.2024.123851","article-title":"Multi-objective optimization and thermodynamic assessment of a solar unit with a novel tube shape equipped with a helical tape","volume":"254","author":"Mohammadzadeh","year":"2024","journal-title":"Appl. Therm. Eng."},{"key":"ref_24","doi-asserted-by":"crossref","first-page":"111766","DOI":"10.1016\/j.est.2024.111766","article-title":"Predicting the performance of a photovoltaic unit via machine learning methods in the existence of finned thermal storage unit","volume":"90","author":"Kiaghadi","year":"2024","journal-title":"J. Energy Storage"},{"key":"ref_25","doi-asserted-by":"crossref","unstructured":"Toks\u00f6z, S.B., I\u015f\u0131k, G., \u015eahin, G., and Akin, E. (2026). Parameter-efficient fine-tuning for photovoltaic cell defect classification: A systematic comparison of LoRA, QLoRA, and full fine-tuning on ConvNeXt-Tiny. Sensors, 26.","DOI":"10.3390\/s26123659"},{"key":"ref_26","doi-asserted-by":"crossref","unstructured":"Qiu, S., Cui, X., Ping, Z., Shan, N., Li, Z., Bao, X., and Xu, X. (2023). Deep Learning Techniques in Intelligent Fault Diagnosis and Prognosis for Industrial Systems: A Review. Sensors, 23.","DOI":"10.3390\/s23031305"},{"key":"ref_27","doi-asserted-by":"crossref","first-page":"126286","DOI":"10.1109\/ACCESS.2021.3110947","article-title":"Deep Learning-Based Fault Diagnosis of Photovoltaic Systems: A Comprehensive Review and Enhancement Prospects","volume":"9","author":"Mansouri","year":"2021","journal-title":"IEEE Access"},{"key":"ref_28","doi-asserted-by":"crossref","first-page":"13852","DOI":"10.1109\/ACCESS.2022.3140287","article-title":"Hybrid Deep Learning Model for Fault Detection and Classification of Grid-Connected Photovoltaic System","volume":"10","author":"Alrifaey","year":"2022","journal-title":"IEEE Access"},{"key":"ref_29","first-page":"10347","article-title":"Training data-efficient image transformers & dis-tillation through attention","volume":"Volume 139","author":"Touvron","year":"2021","journal-title":"Proceedings of the 38th International Conference on Machine Learning (ICML 2021)"},{"key":"ref_30","first-page":"1321","article-title":"On Calibration of Modern Neural Networks","volume":"Volume 70","author":"Guo","year":"2017","journal-title":"Proceedings of the 34th International Conference on Machine Learning (ICML), Sydney, Australia, 6\u201311 August 2017"},{"key":"ref_31","doi-asserted-by":"crossref","first-page":"118049","DOI":"10.1016\/j.eswa.2022.118049","article-title":"Distillation-enhanced fast neural architecture search method for edge-side fault diagnosis of wind turbine gearboxes","volume":"208","author":"Wu","year":"2022","journal-title":"Expert Syst. Appl."},{"key":"ref_32","doi-asserted-by":"crossref","unstructured":"Shan, N., Xu, X., Bao, X., and Qiu, S. (2022). Fast Fault Diagnosis in Industrial Embedded Systems Based on Compressed Sensing and Deep Kernel Extreme Learning Machines. Sensors, 22.","DOI":"10.3390\/s22113997"},{"key":"ref_33","doi-asserted-by":"crossref","first-page":"112245","DOI":"10.1016\/j.solener.2023.112245","article-title":"An efficient CNN-based detector for photovoltaic module cells defect detection in electroluminescence images","volume":"267","author":"Liu","year":"2024","journal-title":"Sol. Energy"},{"key":"ref_34","doi-asserted-by":"crossref","first-page":"20671","DOI":"10.1038\/s41598-024-72019-5","article-title":"A PV cell defect detector combined with transformer and attention mechanism","volume":"14","author":"Lang","year":"2024","journal-title":"Sci. Rep."},{"key":"ref_35","doi-asserted-by":"crossref","unstructured":"Toks\u00f6z, S.B., I\u015f\u0131k, G., \u015eahin, G., and Akin, E. (2026). Evaluating convolutional and transformer architectures for photovoltaic defect classification via electroluminescence imagery. Sensors, 26.","DOI":"10.3390\/s26123775"},{"key":"ref_36","doi-asserted-by":"crossref","first-page":"502","DOI":"10.1016\/j.renene.2021.07.070","article-title":"Automatic Fault Classification in Photovoltaic Modules Using Convolutional Neural Networks","volume":"179","author":"Alves","year":"2021","journal-title":"Renew. Energy"},{"key":"ref_37","doi-asserted-by":"crossref","first-page":"1434","DOI":"10.1049\/rpg2.12383","article-title":"A compressed sensing and CNN-based method for fault diagnosis of photovoltaic inverters in edge computing scenarios","volume":"16","author":"Wang","year":"2022","journal-title":"IET Renew. Power Gener."},{"key":"ref_38","doi-asserted-by":"crossref","unstructured":"Wu, Z., Guo, Y., Lin, W., Yu, S., and Ji, Y. (2018). A Weighted Deep Representation Learning Model for Imbalanced Fault Diag-nosis in Cyber-Physical Systems. Sensors, 18.","DOI":"10.3390\/s18041096"},{"key":"ref_39","doi-asserted-by":"crossref","first-page":"102622","DOI":"10.1016\/j.rineng.2024.102622","article-title":"Deep Learning Approaches for Visual Faults Diagnosis of Photovoltaic Systems: State-of-the-Art Review","volume":"23","author":"Jalal","year":"2024","journal-title":"Results Eng."}],"container-title":["Sensors"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.mdpi.com\/1424-8220\/26\/13\/4256\/pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,13]],"date-time":"2026-07-13T08:08:01Z","timestamp":1783930081000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.mdpi.com\/1424-8220\/26\/13\/4256"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,7,4]]},"references-count":39,"journal-issue":{"issue":"13","published-online":{"date-parts":[[2026,7]]}},"alternative-id":["s26134256"],"URL":"https:\/\/doi.org\/10.3390\/s26134256","relation":{},"ISSN":["1424-8220"],"issn-type":[{"value":"1424-8220","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,7,4]]}}}