Plant Leaf Disease Detection and Pesticide Recommendation System using Deep Learning

Abstract

An Agriculture plays a crucial role in the economy, yet crop productivity is significantly affected by plant leaf diseases that often go undetected at early stages. Farmers, especially in rural areas, face challenges in accurately identifying diseases and selecting appropriate pesticides, leading to reduced yield and increased costs. Existing solutions are either manual, time-consuming, or lack intelligent decision-making capabilities. This paper presents a deep learning-based plant leaf disease detection and pesticide recommendation system designed to address these challenges in an end-to-end manner. The system employs a Convolutional Neural Network (CNN) model trained on a large dataset of plant leaf images to accurately classify diseases across multiple crops. The trained model achieves high accuracy in identifying both healthy and diseased leaves under varied environmental conditions. Once a disease is detected, the system integrates a recommendation module that suggests suitable pesticides and preventive measures based on the identified disease. The complete solution is implemented as a user-friendly web application where users can upload leaf images and receive instant results. The system is designed for real-time usage, ensuring accessibility and ease of use for farmers without requiring technical expertise. By combining computer vision and deep learning with practical agricultural knowledge, this system provides an efficient, scalable, and cost-effective solution for early disease detection and crop management. It has the potential to reduce crop losses, improve productivity, and support sustainable farming practices.

Introduction

The paper presents an AI-based system for detecting plant leaf diseases and recommending suitable pesticides to improve agricultural productivity. Traditional disease identification methods are manual, time-consuming, and often inaccurate, leading to crop losses and improper pesticide use. To address this, the proposed system uses deep learning, particularly Convolutional Neural Networks (CNNs) with transfer learning, to automatically analyze leaf images and identify diseases with high accuracy.

The system follows a client-server architecture with a React-based frontend for image upload and result display, and a FastAPI backend for processing and prediction. Users can upload leaf images, which are analyzed by a trained model (using datasets like PlantVillage), and receive real-time disease detection along with pesticide recommendations and preventive measures.

Experimental results show high accuracy (around 95–98%), though performance may vary based on lighting, background, and image quality. The system is user-friendly, web-based, and designed for farmers, offering a complete solution from detection to treatment guidance.

Overall, the project provides an efficient, accessible, and intelligent tool for crop disease management, with future improvements suggested such as mobile integration, real-time detection, and inclusion of environmental data.

Conclusion

The proposed Plant Leaf Disease Detection and Pesticide Recommendation System presents an automated and intelligent framework for accurate crop disease diagnosis using deep learning techniques. By integrating Convolutional Neural Networks (CNNs) and transfer learning models, the system effectively classifies plant leaf diseases with high accuracy. The incorporation of a rule-based pesticide recommendation module enhances practical applicability by providing dosage guidance and safety precautions. The web-based deployment ensures real-time accessibility, making the system suitable for precision agriculture and decision-support applications. Future work will focus on extending the system to real-field conditions with complex backgrounds, integrating IoT-based environmental monitoring sensors, and developing a mobile application for wider accessibility. Additionally, incorporating severity prediction using advanced segmentation models and implementing multilingual voice support can further enhance usability and farmer adoption.

References

[1] M. Jung et al., \"Construction of deep learning-based disease detection model for tomato, bell pepper and potato using image pairs,\" Scientific Reports (Nature), 2023. Uses CNNs (ResNet50, Efficient Net) on diseased-healthy pairs for multi-crop detection. [2] \"Detection and identification of plant leaf diseases using YOLOv4,\" Frontiers in Plant Science, 2024. Achieves 99% accuracy with YOLOv4 for real-time detection and crop yield protection. [3] \"Advancing plant leaf disease detection integrating machine learning and deep learning,\" Scientific Reports (Nature), 2025. Compares VGG-19, Inception v3 hybrids for multi-crop leaves (e.g., potato, banana) [4] \"A Review on the Detection of Plant Disease Using Machine Learning and Deep Learning,\" PMC, 2025. CNN outperforms traditional ML (98% accuracy); covers mobile deployments. [5] \"Context-Aware Pesticide Recommendation via Few-Shot Pest Detection,\" arXiv, 2026. Lightweight CNN for pest detection tied to sustainable pesticide suggestions. [6] R. Patel et al., \"YOLO-AgriNet: A Deep Learning-Based Model for Real-Time Plant Disease Detection,\" Scientific Research Publishing Journal, vol. 16, no. 8, Aug. 2025. [7] S. Ramachandran, “Indian Sign Language Dataset,” Kaggle, 2021. [8] J. Liu et al., \"A review of plant leaf disease identification by deep learning,\" Frontiers in Plant Science, vol. 16, Aug. 2025 [9] X. Wang et al., \"Ag-YOLO: A Real-Time Low-Cost Detector for Precise Agricultural Object Detection,\" Sensors, vol. 22, no. 1, Dec. 2021. [10] K. Sharma et al., \"Detection and Classification of Diseases in Multi-Crop Leaves using CNN-LSTM Hybrid Model,\" arXiv:2505.00741, May 2025. [11] H. Chen et al., \"Identification of plant leaf diseases by deep learning based on channel attention mechanism,\" Computers and Electronics in Agriculture, vol. 204, Nov. 2022

Copyright

Copyright © 2026 B. Kanaka Durga, C. Vishwanath, A .Vinay Vamsi, B. Hema Sri, D. Srilatha, M. Shankar Prabanjan. 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.