Defect recognition of solar panel in EfficientNet-B3 network based on CBAM attention mechanism

Defect recognition in solar panels is critical to safeguard their performance and efficiency. Traditional image recognition models have limitations in fine-grained defect feature extraction, which affects the accuracy and efficiency of recognition. In this paper, we propose an EfficientNet-B3 network optimization model based on the CBAM attention mechanism, which significantly improves the recognition of tiny defects in solar panels by combining deep learning techniques and attention mechanisms. Experimental results show that our model exhibits high accuracy on both training and validation sets with gradually decreasing loss. The model achieves an accuracy of 95.22% in complex and variable defect categories, which is significantly better than existing baseline models. An in-depth performance evaluation shows that the model has significant advantages in key performance metrics such as precision, recall, and F1 value, demonstrating its effectiveness and adaptability in the solar panel defect recognition task.