Leaf Disease Detection Using Convolutional Neural Network

Abstract

The aim of the project is to develop a Leaf disease detection system using deep learning, specifically Convolutional Neural Networks (CNNs). The project focuses on classifying images of plant leaves into 39 different disease categories by utilizing deep learning, specifically Convolutional Neural Networks (CNNs), for plant disease detection based on leaf images. The dataset consists of 39 classes with a total of 61,486 images, and various augmentation techniques are applied to increase dataset size. The implementation involves PyTorch, with transformations for data augmentation, dataset creation using Image Folder, and a split into training, validation, and testing sets. The CNN model is designed for image classification, using ReLU activation and soft max for the final layer. The training process involves batch gradient descent, and the model achieves an accuracy of 87% on training data, 84% on validation data, and 83% on test data. The key objectives include utilizing image data, implementing data augmentation techniques, creating a dataset, and training a CNN model to accurately predict and classify plant diseases based on input images. The ultimate goal is to provide a tool that can assist in early detection and diagnosis of plant diseases through automated analysis of leaf images

Introduction

I. INTRODUCTION

Agriculture stands as the backbone of our economy, and safeguarding plant health is paramount for optimal crop production. Detecting leaf diseases is crucial for maintaining plant health and ensuring profitable yields. To address this, deep learning techniques are employed to detect diseases in apple, grape, corn, potato, and tomato plants, encompassing a total of 32 different disease types. The dataset comprises 24,000 images of plant leaves, categorized into various labels such as Apple scab, Black rot, Esca, Corn rust, Early blight, and bacterial spot, among others. These images are resized to 256 x 256 pixels and divided into training and testing datasets, totaling 31,119 images across different plant varieties.

A. Background

Productivity of crops are paramount. Traditionally, farmers relied on visual inspection to identify diseased plants, but this method is often subjective and time-consuming. With advancements in technology, particularly in the fields of computer vision and machine learning, automated plant disease detection systems have emerged as a promising solution. These systems leverage sophisticated algorithms to analyze images of plant leaves and accurately identify signs of disease, such as discoloration, lesions, and deformities. By training on large datasets of labeled images, these algorithms can learn to distinguish between healthy and diseased plants with a high level of accuracy. Overall, plant disease detection technology represents a significant advancement in the field of precision agriculture, offering a proactive approach to plant health management and contributing to global food security.

B. Motivation

Identifying and recognizing leaf diseases offers a solution for addressing the significant challenges posed by crop   diseases, leading to increased productivity and profitability in agriculture. This is particularly beneficial for agricultural institutes and biological research organizations. Additionally, the implementation of Convolutional Neural Networks (CNNs) facilitates automation in the detection process, reducing the reliance on manual inspection methods. This not only saves valuable time but also enables large-scale monitoring, thereby facilitating the prompt identification and containment of disease outbreaks. Moreover, the adoption of CNNs in agriculture contributes to the advancement of precision farming practices. By providing farmers with precise information about the areas affected by diseases, CNNs enable a targeted approach to crop management.

C. Aim and Objective

The aim and objective of the Leaf Disease Detection project are to utilize deep learning techniques, specifically Convolutional Neural Networks (CNNs), to develop a system capable of accurately classifying leaf images into 39 different disease categories.

The project aims to provide a solution for farmers and agriculturalists to identify and address plant diseases promptly, thereby improving crop productivity and reducing economic losses. Additionally, by leveraging the PyTorch framework and open-source contributions, the project aims to create a robust and accessible platform for Leaf disease detection, fostering collaboration among developers and researchers to enhance the efficiency and effectiveness of the model over time.

D. Scope

The Leaf Disease Detection project has a broad scope encompassing various aspects of agricultural innovation, technological advancement, collaboration, accessibility, and education. At its core, the project aims to revolutionize agriculture by providing farmers with a reliable tool for early detection and management of Leaf diseases. Leveraging deep learning techniques, particularly Convolutional Neural Networks (CNNs), the project explores cutting-edge technology to address real-world challenges in agriculture. By making the project open-source, it encourages collaboration among developers, researchers, and agricultural experts worldwide, fostering knowledge sharing, innovation, and continuous improvement. The development of a user-friendly web application ensures accessibility to farmers and stakeholders with varying levels of technical expertise, while scalability allows for deployment across different regions and crops. Additionally, the project aims to raise awareness about the importance of plant disease detection and the potential of technology-driven solutions through educational resources and outreach efforts. Overall, the Leaf Disease Detection project holds the promise of enhancing crop yields, reducing economic losses, and contributing to global food security efforts..

E. Methodology

This area includes the implementation steps and methods consisting of creating and deploying identification, classification techniques. For classification, here we use efficient convolutional neural networks (CNN) algorithm. Here we have two folders train and test. Training used for building the system on plant leaves and test consists of testing the system and detecting the accuracy of the work. First of All, import the necessary packages like OpenCV, NumPy, Tensor Flow, tqdm, matplotlib, etc. Here defining the first function for label images. The developed model builds with four classes of classification. Next, create a function for loading the training data. Training data load images from our folder then resize it. Here resizes the image with a resolution of 50*50. After resizing append the images and corresponding labels to the list. Testing data build with same as mentioned.

II. LITERATURE REVIEWS

These research papers shed light on the significant advancements in plant disease detection using deep learning and image processing techniques. They address the critical need to enhance agricultural productivity by effectively identifying and managing crop diseases. By leveraging technologies like Convolutional Neural Networks (CNNs) and deep learning models, researchers aim to automate disease detection processes, thereby enabling timely intervention and minimizing crop losses. The studies explore various approaches, including real-time disease detection in tomato and apple plants, classification of plant leaf diseases using CNNs, and segmentation-based disease identification. Additionally, they emphasize the importance of addressing agricultural challenges through innovative solutions that harness the power of modern technology, ultimately contributing to global food security and sustainability efforts.

The studies outlined in these papers showcase innovative approaches such as real-time disease detection, classification models trained on extensive datasets, and segmentation-based techniques for precise identification. By harnessing the potential of artificial intelligence and computer vision, these efforts contribute to the development of sustainable farming practices and pave the way for a more resilient agricultural sector in the face of evolving environmental and climatic conditions.

III. STUDY REVIEW

This project focuses on the development of a plant disease detection system using deep learning techniques. Utilizing a convolutional neural network (CNN) built on the PyTorch framework, the system classifies leaf images into 39 different categories of plant diseases. The training data is sourced from the Plant Village dataset. To run the project locally, Python 3.8 is required along with the creation of a virtual environment and installation of dependencies listed in the requirements.txt file. Additionally, users need to download the pre-trained model file and follow instructions to set up the Flask app for deployment. The project is open source, inviting contributions from developers to enhance various aspects such as the user interface, deep learning model, and documentation. Testing images are provided for validation, and a blog link offers insights into the methodology used. The deployed web app features sections including the main page, AI engine, results page, supplements/fertilizer store, and contact information. With a focus on CNN-based plant disease detection, this project aims to provide a valuable tool for farmers and researchers to improve crop health management.

A. Proposed System

We are building a neural network model for image classification. this model will be deployed on the android application for live detection of plant leaf disease through an android phone’s camera. The recognition and classification procedures are depicted in Fig. 1

1. The first step is to collect data. We are using the PlantVillage Dataset, which   is   widely   available. This dataset was released by crowd AI.
2. Pre-processing and Augmentation of the collected dataset is done using pre-processing and Image-data generator API by Kera’s.
3. Building CNN(Convolutional Neural Network) Model (Vgg-19 architecture) for classification of various plant diseases.
4. Developed model will be deployed on the Android Application with help of TensorFlow lite

.The proposed system for plant disease detection leverages advanced deep learning techniques, specifically convolutional neural networks (CNNs), to accurately classify leaf images and identify various types of plant diseases. By utilizing a CNN built on the PyTorch framework, the system can effectively analyze the visual features of leaf images and classify them into different disease categories. The system is trained on a comprehensive dataset, such as the Plant Village dataset, which contains a wide range of images representing healthy and diseased plant leaves. Through extensive training and optimization, the CNN learns to recognize patterns and features indicative of specific diseases, enabling it to provide accurate diagnoses.

In addition to its classification capabilities, the proposed system incorporates a user-friendly interface for seamless interaction. This interface allows users, including farmers and agricultural researchers, to easily upload leaf images for analysis and receive instant feedback on the presence of any diseases. The system's web application provides a convenient platform for users to access its functionalities from any device with internet connectivity. Furthermore, the system can be deployed on various platforms, making it accessible to a wide range of users across different locations.

Overall, the proposed system for plant disease detection represents a significant advancement in agricultural technology, offering a reliable and efficient solution for early disease detection and management. By harnessing the power of deep learning and providing a user-friendly interface, the system has the potential to revolutionize crop health monitoring and contribute to improved agricultural productivity and sustainability.

CONVOLUTIONAL NEURAL NETWORK ARCHITECTURE (VGG-19)

A Convolutional Neural Network has three layers: a convolutional layer, a pooling layer, and a fully connected layer. Fig 2 shows all layers together

B.Future Improvements

Future improvements for the Leaf Disease Detection project could focus on enhancing both the technical capabilities and user experience of the system. One area for improvement is the enhancement of model performance, which involves refining and optimizing the convolutional neural network (CNN) model to improve its accuracy and efficiency in classifying leaf images. Expanding the system's disease detection capabilities by increasing the number of disease categories it can recognize is another important aspect. This expansion would involve diversifying the dataset to include a broader range of leaf diseases and training the model accordingly. Implementing real-time image processing capabilities would enable instantaneous disease detection and diagnosis, enhancing the system's utility for farmers. Additionally, enhancing the user interface of the web application to make it more intuitive and visually appealing would improve user engagement. Localization and multi-language support would make the system accessible to users worldwide by providing support for different languages and regional dialects. Developing a mobile application version of the system would enable farmers to easily access and utilize leaf disease detection functionality on their smartphones or tablets. Lastly, fostering a collaborative community of developers, researchers, and agricultural experts would encourage knowledge sharing and collaboration, further advancing the Leaf Disease Detection project. By focusing on these future improvements, the Leaf Disease Detection project can evolve into a more robust and widely adopted tool for farmers, contributing to improved crop health and food security.

Conclusion

A. Drawbacks While the Leaf Disease Detection project holds significant promise, there are certain drawbacks that warrant consideration. Firstly, the reliance on deep learning models, particularly convolutional neural networks (CNNs), may pose challenges in terms of computational resources and training time, especially for large-scale datasets. Additionally, the accuracy of disease classification may be affected by factors such as variability in environmental conditions, lighting, and image quality, potentially leading to misdiagnosis or false positives. Another limitation lies in the dependency on labeled datasets for training the model, which may not always be comprehensive or representative of all possible disease manifestations. Moreover, the project\'s deployment as a web application may limit accessibility for farmers in remote or low-resource areas with limited internet connectivity or technological infrastructure. Furthermore, the system\'s performance may vary across different plant species and disease types, requiring ongoing refinement and adaptation to ensure effectiveness across diverse agricultural contexts. B. Scalability Scalability is a crucial aspect for the Leaf Disease Detection project, as it determines the system\'s ability to handle increasing volumes of data and user demands as it grows. One potential challenge lies in scaling the deep learning model to accommodate larger datasets without compromising performance or increasing computational overhead. This may require optimizing the model architecture, implementing parallel processing techniques, or leveraging cloud-based resources for distributed computing. Additionally, as the system attracts more users, scalability considerations extend to the web application\'s infrastructure, including server capacity, bandwidth, and response times. Implementing load balancing mechanisms, caching strategies, and horizontal scaling approaches can help ensure that the application remains responsive and reliable under heavy traffic. Furthermore, scalability should also encompass the system\'s ability to adapt to evolving agricultural landscapes and emerging disease threats, necessitating continuous updates, improvements, and integration with new data sources and technologies. Overall, prioritizing scalability in the design and development of the Leaf Disease Detection project is essential for its long-term viability and effectiveness in supporting farmers\' needs and addressing plant health challenges on a broader scale.

References

[1] ” R. G. de Luna, E. P. Dadios and A. A. Bandala, \"Automated Image Capturing System for Deep Learning-based Tomato Plant Leaf Disease Detection and Recognition,\" TENCON 2018 - 2018 IEEE Region 10 Conference, Jeju, Korea (South), 2018, pp. 1414-1419, doi: 10.1109/TENCON.2018.8650088. [2] Suma, V. & Shetty, R & Tated, Rishab & Rohan, Sunku & Pujar, Triveni. (2019). CNN based Leaf Disease Identification and Remedy Recommendation System. 395-399. 10.1109/ICECA.2019.8821872. [3] P. Jiang, Y. Chen, B. Liu, D. He and C. Liang, \"Real-Time Detection of Apple Leaf Diseases Using Deep Learning Approach Based on Improved Convolutional Neural Networks,\" in IEEE Access, vol. 7, pp. 59069-59080, 2019, doi: 10.1109/ACCESS.2019.2914929. [4] Geetharamani G., Arun Pandian J., Identification of plant leaf diseases using a nine-layer deep convolutional neural network, Computers & Electrical Engineering, Volume 76, 2019, Pages 323-338, ISSN 0045-7906, https://doi.org/10.1016/j.compeleceng.2019.04.011. [5] P. Soni and R. Chahar, \"A segmentation improved robust PNN model for disease identification in different leaf images,\" 2016 IEEE 1st International Conference on Power Electronics, Intelligent Control and Energy Systems (ICPEICES), Delhi, India, 2016, pp. 1-5, doi: 10.1109/ICPEICES.2016.7853301. [6] Omkar Kulkarni, “Crop Disease Detection using Deep Learning”. 2018 Fourth International Conference on Computing Communication Control and Automation (ICCUBEA) 978-1-5386-5257- 2/18. IEEE,(2018). [7] Abirami Devaraj, Karunya Rathan, Sarvepalli Jaahnavi and K Indira “Identification of Plant Disease using Image Processing Technique” International Conference on Communication and Signal Processing, April 4-6, 2019, India [8] V. Sahithya, B. Saivihari, V. K. Vamsi, P. S. Reddy and K. Balamurugan, \"GUI based Detection of Unhealthy Leaves using Image Processing Techniques,\" 2019 International Conference on Communication and Signal Processing (ICCSP), Chennai, India, 2019, pp. 0818-0822, doi: 10.1109/ICCSP.2019.86979 [9] International Journal of Computer Vision and Image Processing Volume 9 • Issue 1 • January-March 2019 Copyright © 2019, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited. 51 Tomato Plant Leaves Disease Classification Using KNN and PNN Balakrishna K., Maharaja Institute of Technology, Mysore, India Mahesh Rao, Maharaja Institute of Technology, Mysore, India

Copyright

Copyright © 2024 Dr. Sharath M N, Raqeeb Bun Nisa, Muktha M R , Nayana D. 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.