EFM-ResNet: A Feature Enhanced Network for Tobacco Strips Classification

Tobacco strips are the core raw material in tobacco products, and the accuracy of their classification directly impacts product quality and production efficiency. In this paper, we propose a tobacco strip image classification model based on multi-scale fusion attention mechanism and feature enhancement improved ResNet, named EFM-ResNet, to address the issues of detail information loss during feature extraction and the difficulty in capturing long-range dependencies. First, we introduce a multi-scale fusion attention mechanism that enables the model to capture key information in the image, thereby enriching feature details. Second, we develop a feature enhancement module to further improve the feature representation capability, allowing the model to capture more key information. Finally, a WFL loss function is designed to address the limitations of the traditional cross-entropy loss function. Experimental results show that the proposed method achieves an accuracy of 95.87% in the tobacco strips classification task, significantly outperforming traditional methods and existing deep learning models. Therefore, the classification method proposed in this paper can substantially improve tobacco strips classification accuracy and provide strong technical support for the intelligent production of the tobacco industry.