AiKrishi Crop Disease Detection Using Vision Transformer

Abstract

Plant diseases pose a serious threat to global food security and sustainable agriculture, causing significant economic losses and affecting rural livelihoods. Traditional disease detection methods, such as manual inspection and lab tests, often lack efficiency, scalability, and accuracy. This paper explores the use of vision transformers (ViTs), a cutting-edge deep learning technique, to enhance plant disease detection. ViTs utilize self-attention mechanisms to analyze complex patterns in plant images, enabling precise and automated classification. A thorough review of deep learning applications in agriculture highlights the growing interest in ViTs for disease identification. This study also details an effective methodology for training and evaluating ViT models on a well-balanced dataset of 55 plant disease classes. Experimental findings confirm the superior accuracy of ViTs, showcasing their transformative potential in precision agriculture. By improving early disease diagnosis, ViTs can contribute to more sustainable farming practices and increased agricultural productivity.

Introduction

Agriculture is essential for human survival, providing food, fiber, and fuel. Sustainable farming is critical to ensure long-term food security, environmental preservation, and economic stability. A key aspect of sustainable agriculture is effective plant disease detection, which directly influences crop yields and food supply. Traditional disease detection methods—visual inspection and laboratory testing—are often slow, subjective, costly, and not scalable for large farms.

Modern technologies, particularly artificial intelligence (AI) and deep learning, have transformed plant disease identification. Convolutional Neural Networks (CNNs) have improved accuracy but still struggle with complex patterns and require large datasets. Recently, Vision Transformers (ViTs)—originally developed for natural language processing—have been adapted for computer vision and outperform CNNs by using self-attention mechanisms that capture intricate image patterns more effectively and efficiently.

ViTs require fewer computational resources, offer superior accuracy, and have been successfully used in various studies for plant disease classification. Hybrid models combining ViTs and CNNs, as well as optimized versions like GreenViT, further enhance detection accuracy and efficiency. Smartphone applications leveraging ViTs enable real-time disease diagnosis in the field, supporting farmers directly.

The proposed system in the research uses a large, diverse dataset of plant leaf images from multiple species and diseases, applying advanced preprocessing and data augmentation to improve model robustness. The ViT architecture divides images into patches and processes them through transformer encoder layers, utilizing self-attention to recognize global and local features for precise disease classification.

Overall, ViTs represent a promising solution for automated, scalable, and accurate plant disease detection, facilitating precision agriculture, reducing reliance on manual methods, and contributing to sustainable farming and food security.

Conclusion

The integration of AI-based plant disease detection presents a transformative approach to modern agriculture. With the increasing threat of plant diseases affecting global food production, early detection and intervention are crucial. This technology leverages advanced deep learning models to identify and classify plant diseases with high precision, reducing dependency on traditional manual inspections. By providing real-time, accurate results, it enhances decision-making for farmers and agricultural experts. Moreover, the system is cost-effective, scalable, and accessible through mobile applications, making it a valuable tool for farmers of all scales. It significantly reduces pesticide misuse, leading to healthier crops and a more sustainable environment. Additionally, its application extends beyond farms to research, government initiatives, and smart agricultural practices, reinforcing food security and productivity. Despite its advantages, challenges such as model training with diverse datasets and ensuring usability for non-technical users remain. However, continuous advancements in AI, IoT, and cloud computing will further refine these systems. AI-driven plant disease detection is a revolutionary step towards precision agriculture. By improving crop health monitoring and optimizing resource utilization, this technology ensures a sustainable, productive, and secure agricultural future. Widespread adoption will contribute to increased yields, reduced losses, and enhanced global food security.

References

[1] Borhani, A., et al. (2023). Vision Transformers for Real-Time Automated Plant Disease Classification. Computers and Electronics in Agriculture. [2] Zhang, Y., et al. (2022). Deep Learning-Based Computer Vision for Plant Disease Detection. IEEE Access. [3] Sharma, R., et al. (2023). AI-Powered Crop Disease Detection: A Comparative Study of CNNs and Vision Transformers. Agricultural Systems. [4] Lee, J., et al. (2021). Smart Agriculture: Deep Learning for Plant Disease Classification. Sensors. [5] Patel, S., et al. (2023). Advancements in AI-Based Plant Disease Identification: The Role of Vision Transformers. Frontiers in Plant Science. [6] C. Usha Kumari, N. Arun Vignesh, A. K. Panigrahy, L. Ramya, and T. Padma, “Fungal disease in cotton leaf detection and classification using neural networks and support vector machine,” International Journal of Innovative Technology and Exploring Engineering, vol. 8, no. 10, pp. 3664–3667, 2019, doi: 10.35940/ijitee.J9648.0881019. [7] R. Balodi, S. Bisht, A. Ghatak, and K. H. Rao, “Plant disease diagnosis: technological advancements and challenges,” Indian Phytopathology, vol. 70, no. 3, pp. 275–281, 2017, doi: 10.24838/ip.2017.v70.i3.72487. [8] A. J. Mastin, F. Van Den Bosch, F. Van Den Berg, and S. R. Parnell, “Quantifying the hidden costs of imperfect detection for early detection surveillance,” Philosophical Transactions of the Royal Society B: Biological Sciences, vol. 374, no. 1776, 2019, doi: 10.1098/rstb.2018.0261. [9] D. Migliorini, L. Ghelardini, E. Tondini, N. Luchi, and A. Santini, “The potential of symptomless potted plants for carrying invasive soilborne plant pathogens,” Diversity and Distributions, vol. 21, no. 10, pp. 1218–1229, 2015, doi: 10.1111/ddi.12347. [10] C. Temple, A. G. Blouin, S. Tindale, S. Steyer, K. Marechal, and S. Massart, “High throughput sequencing technologies complemented by grower’s perception highlight the impact of tomato virome in diversified vegetable farms,” Frontiers in Sustainable Food Systems, 2023, doi: 10.1101/2023.01.12.523758. [11] S. Padhee and D. Nandan, “Design of automated visual inspection system for beverage industry production line,” Traitement du Signal, vol. 38, no. 2, pp. 461–466, 2021, doi: 10.18280/ts.380225. [12] A. Mhaned, M. Salma, E. H. Mounia, and B. Jamal, “Contribution to smart irrigation based on internet of things and artificial intelligence,” Lecture Notes in Networks and Systems, vol. 625 LNNS, pp. 537–549, 2023, doi: 10.1007/978-3-031-28387-1_45. [13] A. Mhaned, S. Mouatassim, M. El Haji, and J. Benhra, “Low-cost smart irrigation system based on internet of things and fuzzy logic,” Communications in Computer and Information Science, vol. 1677 CCIS, pp. 78–89, 2022, doi: 10.1007/978-3-031-20490-6_7. [14] S. Fouguira, A. Mhaned, M. Ben Abbou, E. Ammar, M. El Haji, and J. Benhra, “Effectiveness of compost use in salt-affected soil in an automated greenhouse irrigation system,” E3S Web of Conferences, vol. 364, 2023, doi: 10.1051/e3sconf/202336403002. [15] R. Sharma et al., “Plant disease diagnosis and image classification using deep learning,” Computers, Materials and Continua, vol. 71, no. 2, pp. 2125–2140, 2022, doi: 10.32604/cmc.2022.020017.

Copyright

Copyright © 2025 Prof. Minakshi Ramteke, Prof. Tarun Yengantiwar , Ms. Vaishali Vijay Gahukar. 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.