An improved YOLOv5 model based on visual attention mechanism: Application to recognition of tomato virus disease

Abstract

• A deep learning model based on attention mechanism is proposed for tomato virus disease recognition. • The recognition accuracy is improved while maintaining the same detection speed. • It provides technical support for other researches related to plant disease recognition. Traditional target detection methods cannot effectively screen key features, which leads to overfitting and produces a model with a weak generalization ability. In this paper, an improved SE-YOLOv5 network model is proposed for the recognition of tomato virus diseases. Images of tomato diseases in greenhouses were collected using a mobile phone, and the collected images were expanded. A squeeze-and-excitation (SE) module was added to a YOLOv5 model to realize the extraction of key features, using a human visual attention mechanism for reference. The trained network model was evaluated on the test set of tomato virus diseases. The accuracy was 91.07%, which was 7.12%, 17.85% and 8.91% higher than that of the Faster regions with convolutional neural network features (R-CNN) model, single-shot multiBox detector (SSD) model and YOLOv5 model, respectively. Meanwhile, the mean average precision (mAP @0.5 ) was 94.10%, which was 1.23%, 16.77% and 1.78% higher than that of the Faster R-CNN model, SSD model and YOLOv5 model. The proposed SE-YOLOv5 model can effectively detect regions of tomato virus disease, which provides disease identification and control theoretical research and technical support.