Tomato leaf diseases recognition model based on improved MobileVit

In order to solve the problem that convolutional neural network is not effective in classifying tomato leaf diseases, a novel method for tomato leaf diseases recognition based on improved MobileViT network was proposed. Firstly, the Squeeze and Excitation module is used to optimize the local and global feature fusion. Secondly, the Global Attention Mechanism (GAM) is introduced after Layer1 and Layer2 of the model to enhance feature representation by capturing global context information, thus improving the expressiveness and robustness of the model. Finally, the ReLU6 activation function is replaced by Mish activation function, which further improves the accuracy of the model. The experimental results show that the improved MobileViT model can well recognize tomato leaf diseases, with an average recognition accuracy of 88.86%, 4.28 percentage points higher than that of the original model. Compared with other convolutional neural network models, the recognition accuracy is higher.