Bundled Tobacco Leaves Vein Detection Based on YOLOV8 and Transfer Learning

The integrity of tobacco leaves is essential for the quality of cigarettes and other tobacco products. Traditional mechanical processing often damages the leaves, causing fragments and broken stems that compromise quality. Our study utilizes the YOLO algorithm and transfer learning to detect and extract veins, preserving leaf integrity effectively. We initially focused on individual tobacco leaves and then expanded to quickly identifying the veins of bundled tobacco leaves. Compared to traditional models, our algorithm shows approximately a 10% improvement in MAP50 and precision. These findings have significant implications for the tobacco leaf de-stemming process, offering innovative solutions for advancing automation and intelligence in the tobacco industry.