Tomato leaf disease detection based on improved YOLOv8

Accurate initial identification of plant diseases is an important safeguard to minimize economic losses of crops. With the application of artificial intelligence technology in agriculture, to improve the accuracy of tomato leaf disease detection, this paper proposes a new tomato leaf disease detection algorithm based on improved YOLOv8 deep neural network. The method adopts a multi-head self-attention mechanism to make the model more focused on the target location and improve the accuracy of tomato leaf disease detection. A deformable convolution method is used to flexibly adjust the sampling position to the input feature map during the convolution operation, so as to better adapt to the changes. In this paper, a total of 9 types of tomato leaf disease images publicly available on Roboflow platform are used as the detection data for experimental validation. The experimental results show that the proposed method significantly improves the average detection accuracy and the operation speed, and the size of the model is significantly reduced. It will meet the requirements of accuracy and timeliness of tomato leaf disease detection, and will provide a theoretical basis for the development of plant leaf disease detection in lightweight intelligent terminals.