TobaccoNet: A deep learning approach for tobacco leaves maturity identification

The maturity of fresh tobacco leaves significantly impacts the quality of subsequent curing processes, making accurate recognition of tobacco leaves maturity crucial. The development of computer vision has facilitated the recognition of tobacco leaves maturity. However, the subtle variations in color and texture among different maturity levels pose a major challenge in effectively identifying various types of tobacco leaves maturity. In this study, tobacco leaves of different maturity levels were collected from tobacco fields for analysis. We propose an innovative and effective deep learning model called TobaccoNet to improve the accuracy of tobacco leaves maturity recognition. The model utilizes ResNet-34 as the backbone network and uniformly segments and recombines preprocessed tobacco leaf image data using a random jigsaw generator. Furthermore, a progressive training approach is employed to train multi-granularity image data, enabling accurate recognition of tobacco leaves maturity. The results demonstrate that the classification accuracy of the TobaccoNet model reaches 96.67 %, highlighting the practicality of our technology in tobacco leaves maturity classification.