Tomato Leaf Disease Detection Using Deep Learning

Abstract

This project will build a machine learning solution for Detection of common diseases of tomatoes from a strong dataset of leaf photographs. We train deep models such as Convolutional from Kaggle. CNNs, MobileNet, and ResNet are employed for leaf condition classification. The system detects diseases such as Tomato Mosaic Virus, Target Spot, Early Blight, and Late Blight, Bacterial Spot and Septoria Leaf Spot with healthy leaves. The backend, Pythonbased, processes the user images and offers real-time Predictions. The frontend, which is built with HTML, CSS, and JavaScript, is intuitive. Interface that allows farmers and gardeners to monitor and diagnose easily tomato plant health, increasing crop management and yield.

Introduction

Overview

Tomato plants are highly susceptible to various diseases that significantly reduce crop yield and quality. Early and accurate disease detection is crucial to prevent large-scale damage and support effective crop management. Traditional manual inspection methods are slow, error-prone, and inconsistent. To address these issues, this project develops an automated tomato leaf disease identification system using deep learning techniques, specifically Convolutional Neural Networks (CNNs) combined with MobileNet and ResNet architectures.

Key Features of the System

• Technology: Uses CNN, MobileNet, and ResNet models for image classification of tomato leaf diseases.

• Diseases Covered: Tomato Mosaic Virus, Early Blight, Late Blight, Target Spot, Bacterial Spot, plus healthy leaves.

• Implementation: Backend built in Python for image processing and prediction; frontend developed with HTML, CSS, and JavaScript for user-friendly interaction.

• Functionality: Users upload leaf images and receive instant disease classification results.

Literature Survey

• Deep learning, particularly CNNs, have revolutionized plant disease detection with high accuracy (over 98%).

• MobileNet is ideal for resource-limited devices due to its lightweight design.

• ResNet improves training of very deep networks via skip connections, enhancing accuracy.

• Studies confirm ResNet is top-performing, but MobileNet is better for real-time, mobile applications.

• Advances in these models have led to scalable, deployable systems aiding sustainable agriculture.

Methodology

• Data Preprocessing: Image resizing, normalization, augmentation (rotation, flipping, brightness adjustment), and dataset splitting into training, validation, and testing.

• Model Training: Employs CNN, MobileNet, and ResNet with transfer learning and fine-tuning to optimize accuracy.

• Backend: Hosts trained models, processes user images, and returns disease predictions via API.

• Frontend: Simple interface allowing image upload and displaying classification results.

• Deployment: Uses cloud platforms (AWS, Heroku, Google Cloud) for scalable access.

Algorithms Explained

• ResNet: Deep neural network with skip connections to prevent vanishing gradients, enabling training of very deep layers and improving classification accuracy.

• CNN: Automatically extracts image features like edges and shapes without manual intervention, enhancing precision in classification tasks.

• MobileNet: Lightweight CNN optimized for mobile/edge devices using depthwise separable convolutions to reduce computational cost while maintaining good accuracy.

Dataset

Includes over 1,000 images per class for diseases like Late Blight, Early Blight, Septoria Leaf Spot, Yellow Leaf Curl Virus, Bacterial Spot, Target Spot, Mosaic Virus, Leaf Mold, and Spider Mites, along with healthy leaves.

Evaluation Metrics

Models are assessed using accuracy, precision, recall, and F1-score, calculated based on true/false positives and negatives.

Results & Analysis

• CNN-based models effectively extract features such as color changes, spots, and leaf deformations.

• Both ResNet and MobileNet models are evaluated via confusion matrices.

• ResNet improves CNN performance with deeper layers.

• MobileNet offers higher accuracy and better performance than ResNet, likely due to its efficient architecture.

Conclusion

This project effectively deploys a deep learning-based system for the detection of tomato leaf diseases using CNN, MobileNet, and ResNet architectures. Based on an image dataset of tomatoes leaf, the system effectively diagnoses tomato leaf diseases like Tomato Mosaic Virus, Tomato Target Spot, Tomato Early blight, Tomato Late blight, Bacterial Spot, and Septoria Leaf Spot, among others. The Python-based backend facilitates effective model inference, while the HTML, CSS, and JavaScript frontend offers a user-friendly interface for easy disease diagnosis. The system suggested works to overcome the various shortcomings and drawbacks of the conventional disease detection approaches by coming up with a solution that offers significantly quicker, more precise, and scalable forms of disease detection. The solution offers a way to significantly aid farmers, crop scientists, and plant health specialists by allowing them to make on-time and sound judgments that will be able to avert loss of crops and ultimately enhance the quality of the overall yield in the long run. Future progress can involve integrating real-time detection functionality through mobile applications, along with integrating the system with IoT-based monitoring systems for agriculture, giving rise to an even stronger and automated tool of effectively controlling disease. VII. ACKNOWLEDGMENT We express our gratitude to all the facilitators who assisted us in completing this research successfully. We express our sincere gratitude to Sreenivasa Institute of Technology and Management Studies (SITAMS) for the facilities and assistance extended.Weare particularly grateful to Dr. R. KaruniaKrishnapriya and Mr. A. Venkatesan for their technical insights and advice regarding hepatocellular carcinoma and machine learning, which significantly enhanced the quality and scope of this work.We also thank our peers and mentors for their encouragement and support throughout the process when we carried out our research. Their direct and indirect comments were of great help to us.

References

[1] Mohanty, S. P., Hughes, D. P., &Salathé, M. (2016). Using Deep Learning for Image-Based Plant Disease Detection. Frontiers in Plant Science, 7, 1419. Available at: https://www.frontiersin.org/journals/plantscience/articles/10.3389/fpls.2016.01419/full [2] Ferentinos, K. P. (2018). Deep Learning Models for Plant Disease Detection and Diagnosis. Computers and Electronics in Agriculture, 145, 311–318. Available at: https://www.sciencedirect.com/science/article/pii/S0168169917315750 [3] He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep Residual Learning for Image Recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 770–778). Available at: https://ieeexplore.ieee.org/document/7780459 [4] Too, E. C., Yujian, L., Njuki, S., &Yingchun, L. (2019). A Comparative Study of Fine-Tuning Deep Learning Models for Plant Disease Identification. Computers and Electronics in Agriculture, 161, 272–279. Available at: https://www.sciencedirect.com/science/article/pii/S0168169919304151 [5] Zhang, S., Huang, W., & Zhang, C. (2020). Plant Disease Recognition Using Convolutional Neural Networks. Plant Methods, 16, 118. Available at: https://plantmethods.biomedcentral.com/articles/10.1186/s13007-020-00657-4 [6] Shoaib, M., Hussain, T., Shah, B., Ullah, I., Shah, S. M., Ali, F., & Park, S. H. (2022). Deep learning-based segmentation and classification of leaf images for detection of tomato plant disease. Frontiers in Plant Science, 13, 1031748. Available at: https://www.frontiersin.org/articles/10.3389/fpls.2022.1031748/full [7] Seno, S., Seno, E. S., & Seno, A. (2024). Bayesian optimized multimodal deep hybrid learning approach for tomato leaf disease detection. Scientific Reports, 14, 72237. Available at: https://www.nature.com/articles/s41598-024-72237-x [8] Zhang, Y., Li, X., & Wang, J. (2024). Tomato leaf disease detection based on attention mechanism and multi-scale feature fusion. Frontiers in Plant Science, 15, 1382802. Available at: https://www.frontiersin.org/articles/10.3389/fpls.2024.1382802/full [9] Khan, A., Nawaz, U., Kshetrimayum, L., Seneviratne, L., & Hussain, I. (2023). Early and Accurate Detection of Tomato Leaf Diseases Using TomFormer. arXiv preprint arXiv:2312.16331. Available at: https://arxiv.org/abs/2312.16331 [10] Ahmed, S., Hasan, M. B., Ahmed, T., Sony, R. K., & Kabir, M. H. (2021). Less is More: Lighter and Faster Deep Neural Architecture for Tomato Leaf Disease Classification. arXiv preprint arXiv:2109.02394. Available at: https://arxiv.org/abs/2109.02394 [11] Agarwal, M., Singh, A., Arjaria, S., Sinha, A., & Gupta, S. (2020). ToLeD: Tomato Leaf Disease Detection using Convolution Neural Network. Procedia Computer Science, 167, 293-302. Available at: https://doi.org/10.1016/j.procs.2020.03.225 [12] Chen, Z., Wang, G., Lv, T., & Zhang, X. (2024). Using a Hybrid Convolutional Neural Network with a Transformer Model for Tomato Leaf Disease Detection. Agronomy, 14(4), 673. Available at: https://doi.org/10.3390/agronomy14040673

Copyright

Copyright © 2025 Ellichetty Bhavana, B Ramya Sree, D Bhanu Priya, Mr. A Venkatesan, Dr. R Karunia Krishnapriya, Mr. Pandreti Praveen, Mr. N. Vijaya Kumar, Mr. V. Shaik Mohammad Shahil. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.