Intelligent Leaf Disease Detection Using Deep Learning: Trends, Challenges, and Innovations

Abstract

Timely detection and classification of plant diseases are essential for effective crop management and ensuring food security. Recent advances in deep learning, particularly Convolutional Neural Networks (CNNs), have enabled automated disease diagnosis using leaf images with high accuracy. This paper reviews research from the past five years, covering model architectures, feature extraction methods, datasets, data augmentation techniques, and evaluation metrics used for disease classification across various crops. Although these approaches perform well in controlled environments, challenges such as varying lighting conditions, complex backgrounds, and poor image quality limit their real-world applicability. Existing systems also lack disease severity estimation and struggle with generalization across diverse crops. Future research directions include multi-crop disease classification, real-time deployment on smart devices, and improving model interpretability for practical agricultural use.

Introduction

The text discusses the importance of crop disease detection in agriculture, particularly in Karnataka, a major agricultural state in India. Crop diseases significantly reduce crop quality and yield, leading to economic losses and increased use of pesticides and chemicals, which negatively affect human health, biodiversity, and the environment. Therefore, early and accurate disease detection is essential for effective crop management and reducing harmful chemical usage.

Traditional disease detection methods rely on visual inspection, laboratory testing, microscopy, and consultation with experts. Although effective, these methods are time-consuming, expensive, require expertise, and are not suitable for large-scale farming. To overcome these limitations, researchers are increasingly using computer vision and machine learning technologies, especially deep learning models such as Convolutional Neural Networks (CNNs), for automatic plant disease detection using leaf images.

Leaf images are commonly used because disease symptoms like discoloration, spots, lesions, and wilting often appear first on leaves, making them reliable indicators of plant health. Publicly available datasets such as PlantVillage and others have supported research in this field. However, many datasets are captured in controlled environments with clear images, while real-world field conditions involve poor lighting, shadows, overlapping leaves, and complex backgrounds, which reduce model accuracy.

The paper highlights another major limitation in existing systems: most models only classify leaves as healthy or diseased and do not estimate disease severity. Severity estimation is important for determining treatment urgency and appropriate pesticide application. Although some recent studies attempt severity analysis, achieving accurate severity prediction remains challenging.

The survey reviews more than 15 research papers on plant disease detection and discusses:

• Commonly used datasets
• Different crops considered for disease detection
• Tools and frameworks used for model development
• Challenges in real-world deployment
• Future research directions

The reviewed datasets include PlantVillage, BananaLSD, Rice Leaf Dataset, Tomato Leaf Dataset, Cotton Disease Dataset, and several custom field datasets. Researchers studied diseases across a wide range of crops, including:

• Food crops: rice, wheat, maize, millets, pulses
• Commercial crops: cotton, sugarcane, groundnut
• Plantation crops: tea, coffee
• Vegetables: tomato, potato, chilli, cassava
• Fruits: mango, grapes, apple, citrus
• Spices: pepper and cardamom

The research commonly uses tools and frameworks such as PyTorch, TensorFlow, Keras, and Scikit-learn, with GPU acceleration using CUDA and RTX GPUs for efficient deep learning computation.

Conclusion

Deep learning methods, especially Convolutional Neural Networks (CNNs), have significantly contributed to plant disease detection and classification using leaf images. These models achieve high accuracy across different crops, supporting precision agriculture. However, their performance is often limited in real-world conditions due to environmental variations, noisy data, and the absence of disease severity assessment. Although considerable progress has been made, further improvements in model generalization, diverse datasets, and real-time deployment are essential to make these solutions more practical and accessible for farmers worldwide.

References

[1] R. Rashid, W. Aslam, R. Aziz, and G. Aldehim, “An early and smart detection of corn plant leaf diseases using IoT and deep learning multi-models,” IEEE Access, vol. 12, pp. 23149–23162, 2024. [2] H. Yu, J. Liu, C. Chen, A. A. Heidari, Q. Zhang, H. Chen, M. Mafarja, and H. Turabieh, “Corn leaf diseases diagnosis based on K-means clustering and deep learning,” IEEE Access, vol. 9, pp. 143824–143835, 2021. [3] T. Daniya, C. Vidyadhari, and S. Aluri, “Rice plant disease detection using sensing recognition strategy based on artificial intelligence,” Journal of Mobile Multimedia, vol. 18, no. 3, pp. 705–722, 2022. [4] Z. Zinonos, S. Gkelios, A. F. Khalifeh, D. G. Hadjimitsis, Y. S. Boutalis, and S. A. Chatzichristofis, “Grape leaf diseases identification system using convolutional neural networks and LoRa technology,” IEEE Access, vol. 10, pp. 122–133, 2021. [5] B. R. Reddy, G. Kalnoor, M. Devashish, and P. S. K. Reddy, “Deep learning based mobile application for automated plant disease detection,” IEEE Access, 2025. [6] E. Bates, M. Popovi?, C. Marsh, R. Clark, M. Kovac, and B. B. Kocer, “Leaf-level ash dieback disease detection and online severity estimation with UAVs,” IEEE Access, 2025. [7] M. Masood, M. Nawaz, T. Nazir, A. Javed, R. Alkanhel, H. Elmannai, S. Dhahbi, and S. Bourouis, “MaizeNet: A deep learning approach for effective recognition of maize plant leaf diseases,” IEEE Access, vol. 11, pp. 52862–52876, 2023. [8] S. Mumtaz, M. Alshehri, Y. AlQahtani, A. Alshahrani, B. Alabdullah, H. F. Alhasson, and H. Liu, “Advanced leaf classification using multi-layer perceptron for smart crop management,” IEEE Access, 2025. [9] S. Mumtaz, S. Algamdi, H. F. Alhasson, D. A. AlHammadi, A. Jalal, and H. Liu, “Leaf classification for sustainable agriculture and in-depth species analysis,” IEEE Access, 2025. [10] G. G. Rajput and V. B. Doddamani, “Exploring support vector machine and k-nearest neighbors for pigeonpea leaf image disease detection and classification,” in Proc. 2024 4th Asian Conf. Innovation in Technology (ASIANCON), pp. 1–5, 2024. [11] R. Gaithond and V. Murthy, “Exploring deep learning technique for detection of sterility mosaic disease in pigeon pea,” J. Electrical Systems, pp. 20–9, 2024. [12] G. G. Rajput and V. B. Doddamani, “A hybrid system for pigeon pea leaf disease detection and classification (PLDDC) using CNN and SVM,” J. Electrical Systems, vol. 20, no. 3, pp. 5710–5717, 2024. [13] M. Masood, M. Nawaz, T. Nazir, A. Javed, R. Alkanhel, H. Elmannai, S. Dhahbi, and S. Bourouis, “MaizeNet: A deep learning approach for effective recognition of maize plant leaf diseases,” IEEE Access, vol. 11, pp. 52862–52876, 2023. [14] U. Shafi, R. Mumtaz, M. D. M. Qureshi, Z. Mahmood, S. K. Tanveer, I. Ul Haq, and S. M. H. Zaidi, “Embedded AI for wheat yellow rust infection type classification,” IEEE Access, vol. 11, pp. 23726–23738, 2023. [15] A. Salam, M. Naznine, N. Jahan, E. Nahid, M. Nahiduzzaman, and M. E. H. Chowdhury, “Mulberry leaf disease detection using CNN-based smart Android application,” IEEE Access, vol. 12, pp. 83575–83588, 2024. [16] B. R. Reddy, G. Kalnoor, M. Devashish, and P. S. K. Reddy, “Deep learning based mobile application for automated plant disease detection,” IEEE Access, 2025.

Copyright

Copyright © 2026 Mrs. Ashwini S P, Dr. Chethan L S. 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.