City Research Online

Abstract

Automated inspection of photovoltaic (PV) modules plays a critical role in ensuring long-term system reliability and energy yield. In practice, however, the deployment of deep learning–based defect detection systems is often constrained by the limited availability of labeled inspection data. To address this challenge, we propose a novel Self-Supervised Hybrid CNN–Transformer Framework (S²HCT) for data-efficient defect detection in solar panels. Our framework leverages a hybrid architecture that combines the local feature extraction capabilities of Convolutional Neural Networks (CNNs) with the global attention mechanism of Transformers. We employ a self-supervised pre-training strategy based on contrastive learning (SimCLR) to learn robust representations from unlabeled data. The pre-trained model is then fine-tuned on a small labeled dataset for the defect detection task. The effectiveness of the proposed framework is validated on two publicly available benchmark datasets, namely the ELPV dataset and the Solar Panel Defect Detection (SPDD) dataset. Experimental results show that the proposed method consistently outperforms conventional CNN-based models and recent Transformer-based approaches, achieving classification accuracies of 98.5% and 97.9% on the respective datasets. Data efficiency analysis demonstrates that the proposed model maintains high performance even with limited labeled data, achieving 90.3% accuracy with only 10% of the training data. In addition to quantitative evaluation, attention map visualizations are presented to illustrate how the model focuses on defect-relevant regions, providing insight into its decision-making process. These results indicate that the proposed framework offers a practical and effective solution for data-efficient defect detection in photovoltaic inspection and related industrial applications.

Publication Type:	Article
Additional Information:	© 2026. This manuscript version is made available under the CC-BY-NC-ND 4.0 license https://creativecommons.org/licenses/by-nc-nd/4.0/
Publisher Keywords:	solar panel inspection, defect detection, deep learning, self-supervised learning, CNN-Transformer hybrid, data efficiency, computer vision
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering
Departments:	School of Science & Technology School of Science & Technology > Department of Engineering
SWORD Depositor:	Symplectic Administrator

Export

Downloads