Intelligent Cotton Leaf Disease Detection System using Deep Learning and Multi-Layer Validation Framework

Abstract

Crop diseases in cotton pose a significant economic threat to agriculture across the globe. Most of the diseases can cause a reduction in crop yield of up to 30–40%. For the agriculture sector to thrive, a robust system is required to diagnose and treat crop diseases efficiently. Current approaches rely on manual approaches and are time-consuming, error-prone and require the specialist expertise of agronomists. In this paper, we present a full-stack webbased system that includes deep learning classification for disease detection and employs a novel multi-layer validation system. The classification system utilises a deep Convolutional Neural Network and achieves up to 87.5% classification accuracy across ten diseases including the difficult to diagnose Fusarium Wilt, Bacterial Blight and aphids infestation. Our system employs a dual-validation system where rule-based image analysis is first used to verify if an image is of a cotton leaf, followed by AI-powered verification of the disease. Our system achieves 98% rejection accuracy for non-cotton images and 95% acceptance accuracy for real cotton leaf images. It includes a Flask-based REST API backend that utilises strict validation protocols and evaluates 16 distinct parameters from an image, including green color dominance, skin tone detection, edge density and texture uniformity. A bilingual treatment recommendation system provides on-the-spot treatment advice in both English and Hindi. The end-to-end system requires no specialist hardware, enabling specialist disease diagnosis to be carried out on a web browser, making it accessible to farmers worldwide.

Introduction

The text presents a deep learning-based cotton disease detection system designed to improve agricultural diagnosis and reduce crop losses caused by pests and pathogens. Cotton is a major global crop, but traditional disease identification methods rely on manual inspection, which is slow, subjective, and impractical for large-scale farming. Existing AI-based systems also face issues such as high false detections, poor real-world performance, and lack of robustness.

To address these challenges, the proposed work introduces a comprehensive cotton leaf disease detection framework with several key contributions: a dual-layer validation system to filter non-cotton images, a deep CNN model trained to classify 10 cotton diseases with improved accuracy, a real-time web-based diagnosis platform, and a bilingual treatment recommendation system (English and Hindi). The system also uses advanced image feature analysis such as color, texture, edge, and shape properties.

The literature review shows that while deep learning models like CNNs, VGG, and ResNet achieve high accuracy in controlled datasets, they often fail in real-world conditions due to poor generalization and lack of validation. Cotton-specific studies also suffer from limited datasets and weak field testing. Additionally, most existing systems lack proper image validation, species identification, and integrated treatment recommendations.

The proposed methodology uses a three-tier architecture (frontend, Flask backend, and model/data layer). A key innovation is a dual-layer validation framework. The first layer applies rule-based checks using 16 image parameters such as color space analysis, texture, edge density, resolution, and skin detection to reject non-cotton images. The second layer uses a CNN-based AI classifier (VGG16-based transfer learning) to confirm cotton leaf authenticity before disease classification.

Conclusion

This research presents a comprehensive cotton disease detection system that addresses critical gaps in agricultural AI applications through innovative dual-layer validation, robust deep learning classification, and accessible web-based deployment. The system achieves 98% accuracy in filtering noncotton images while maintaining 87.5% disease classification accuracy across 10 disease categories. Key contributions include the novel multi-parameter validation framework combining rule-based and AI approaches, production-ready full-stack implementation with RESTful API architecture, bilingual treatment recommendation system, and demonstrated real-world usability in farming communities. Future research directions include: integration of temporal disease progression tracking through multiple image uploads, expansion to additional cotton varieties and geographical regions, development of mobile applications for offline operation, incorporation of environmental data (temperature, humidity) for enhanced prediction, and implementation of federated learning for continuous model improvement while preserving farmer privacy. The system demonstrates that practical, accessible AI solutions for agriculture require careful attention to validation, deployment infrastructure, and end-user needs beyond pure classification accuracy. This holistic approach provides a blueprint for developing impactful agricultural AI systems.

References

[1] R. Kumar and A. Singh, “Economic impact assessment of cotton diseases in major producing regions,” Agricultural Economics Review, vol. 45, no. 3, pp. 234-251, 2023. [2] S. P. Mohanty, D. P. Hughes, and M. Salathe, “Using deep learning for´ image-based plant disease detection,” Frontiers in Plant Science, vol. 7, article 1419, 2016. [3] K. P. Ferentinos, “Deep learning models for plant disease detection and diagnosis,” Computers and Electronics in Agriculture, vol. 145, pp. 311318, 2018. [4] X. Xie, Y. Ma, B. Liu, J. He, S. Li, and H. Wang, “A deep-learning-based real-time detector for grape leaf diseases using improved convolutional neural networks,” Frontiers in Plant Science, vol. 11, article 751, 2020. [5] S. Ramesh and D. Vydeki, “Recognition and classification of paddy leaf diseases using optimized deep neural networks with Jaya algorithm,” Information Processing in Agriculture, vol. 7, no. 2, pp. 249-260, 2020. [6] L. Zhang, L. Zhang, and A. C. Bovik, “A feature-enriched completely blind image quality evaluator,” IEEE Transactions on Image Processing, vol. 24, no. 8, pp. 2579-2591, 2015. [7] V. Kumar, A. Arora, and P. Singh, “Automated validation framework for plant leaf images using machine learning,” Expert Systems with Applications, vol. 168, article 114334, 2021. [8] J. Amara, B. Bouaziz, and A. Algergawy, “Deep learning-based approach for banana leaf disease classification,” Data in Brief, vol. 25, 2019. [9] P. Too, L. Yujian, S. Njuki, and L. Yingchun, “A comparative study of fine-tuned deep learning models for plant disease identification,”Computers and Electronics in Agriculture, vol. 161, pp. 272–279, 2019. [10] S. Sladojevic, M. Arsenovic, A. Anderla, D. Culibrk, and D. Stefanovic, “Deep neural networks based recognition of plant diseases by leaf image classification,” Computational Intelligence and Neuroscience, 2016.

Copyright

Copyright © 2026 Prof. Krunal Mudafale, Dhanshree Nanotkar, Sanchita Chaudhari, Sanjana Motghare, Vaidehi Athale. 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.