Title: Tomato disease classification and recognition using machine learning

Authors: Bingjie Liu; Vladimir Mariano

Addresses: College of Computing and Information Technologies, National University, Manila 1008, Philippines; School of Information Engineering, Jiangxi V&T College of Communications Nanchang, Jiangxi 330013, China ' College of Computing and Information Technologies, National University, Manila 1008, Philippines

Abstract: Tomatoes are one of the primary vegetable foods in human society, yet they are susceptible to various diseases. This study used neural network algorithms to design an image recognition system for tomato leaf diseases. Based on the PlantVillage tomato dataset, the image data were preprocessed and normalised, and training and testing sets were selected and sent to three models: convolutional neural network (CNN), residual neural network (ResNet), and visual geometry group (VGG), for comparative training analysis. The experiment employed the cross-entropy function to balance sample differences, and the model training process was enhanced through reasonable parameter settings. Furthermore, the deployment challenges and solutions of the experimental model in real-world scenarios were discussed. After training, the VGG model performed the best, achieving a recognition accuracy rate of 84% and a precision rate of 83.5%. This result demonstrates that with reasonable hyperparameter settings and model configuration, the VGG model has significant advantages in recognising tomato leaf diseases and can provide powerful technical support for agricultural production.

Keywords: tomato disease classification; convolutional neural network; CNN; visual geometry group; VGG; cross entropy.

DOI: 10.1504/IJSN.2025.150112

International Journal of Security and Networks, 2025 Vol.20 No.4, pp.221 - 230

Received: 07 Nov 2024
Accepted: 20 Nov 2024
Published online: 01 Dec 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article