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

Abstract

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

Tea leaf disease detection and identification based on YOLOv7 (YOLO-T)

“…Except for the training time, the result of a recent study 55 that also used the YOLOv5 version for the tea leaf disease was consistent with those of the present research. YOLOv5 employs a Focus structure that requires less www.nature.com/scientificreports/ Compute Unified Device Architecture (CUDA) memory, a reduced layer, and enhanced forward and backpropagation.…”

“…Tea disease targets are often small, and the growing area's complex background easily impedes the procedure of their smart detection. In addition, several tea diseases are concentrated throughout the entire leaf surface, necessitating inferences from global data 55 . Regarding the leaf disease detection task, where accuracy was the essential factor, the proposed YOLO-T model is superior to other models.…”

An Improved YOLOv5 for Accurate Detection and Localization of Tomato and Pepper Leaf Diseases

“…The obtained results showed that the proposed method achieved an mAP of 98.2%. Lin et al [17] proposed a real-time detection model called TSBA-YOLO for tea disease detection using a self-generated dataset. The model incorporates the self-attention mechanisms of convolutional and transformer layers to enhance global perception and obtain more contextual information.…”

A tomato disease identification method based on leaf image automatic labeling algorithm and improved YOLOv5 model

“…where W i and W j are learnable weights that are not < 0; ε = 0.0001 makes the denominator always > 0, ensuring the stability of the value. Lin et al 31 used BiFPN to replace the multi-scale feature fusion network of YOLOv5, so that it could better integrate multi-scale features of tea diseases and enhance the expression ability of the subtle features of tea diseases. In the present study, we use the BiFPN structure to replace the PANet structure of the third Concat layer in the Neck of the network and add the output features from layer six to the input channels of this Concat layer to improve the fusion ability of the critical features of various diseases such as stripes of the diseased spots and leaf morphology, such that the model has a more balanced identification effect on various diseases.…”