TSBA-YOLO: An Improved Tea Diseases Detection Model Based on Attention Mechanisms and Feature Fusion

Tea diseases have a significant impact on the yield and quality of tea during growth trees. The shape scale are variable, disease targets usually small, with intelligent detection processes also easily disturbed by complex background growing region. In addition, some concentrated in entire area leaves, needing to be inferred from global information. Common target models difficult solve these problems. Therefore, we proposed an improved model called TSBA-YOLO. We use dataset collected at Maoshan Factory China. self-attention mechanism was used enhance ability obtain information diseases. BiFPN feature fusion network adaptively spatial (ASFF) technology were improve multiscale resist interference. integrated Shuffle Attention problem identifications small-target data-enhancement methods transfer learning expand relocate parameters learned other plant datasets detection. Finally, SIoU further accuracy regression. experimental results show that is good solving series problems encountered recognition ahead mainstream models, speed reaches real-time level.