Cotton Leaf Disease Detection and Classification Using Deep Learning A Comprehensive Critical Review

Deep learning has substantially advanced automated cotton leaf disease recognition. CNNs and transfer-learning architectures such as VGG, ResNet, Inception, Xception, DenseNet, EfficientNet and MobileNet have produced strong results in the reviewed literature [3], [11], [12], [14], [16], [19], [31], [36]. Hybrid and ensemble models can improve complementary feature representation [1], [11], [22], while attention and XAI increase interpretability [2], [23], [24]. Recent work is also moving toward YOLO-based localization [4], [5], [29], lightweight inference [10], [15], [20], [25], field datasets [6], [9], cross-dataset evaluation and distributed learning [7]. However, reported accuracy values must be interpreted cautiously. Differences in dataset composition, augmentation, class definitions, validation protocols and environmental conditions make direct numerical ranking scientifically inappropriate [33], [34]. The major unresolved challenges are dataset diversity, field robustness, leakage-safe evaluation, early-stage detection, severity estimation, explanation validity, computational efficiency and adaptation over time. The future of cotton disease detection should therefore move from isolated accuracy-oriented classifiers toward complete intelligent agricultural decision-support systems that can detect, localize, classify, quantify and explain disease under real field conditions. The proposed multi-stage architecture provides a practical research pathway toward that goal.

[1]   S. Kanakala, R. Ch. Manoranjini, P. Ravi, M. Riyazuddin, and K. Naresh, “Hybrid deep learning approach for cotton leaf disease detection and management using fine-tuned VGG16 and Inception v3 models,” International Journal of Systematic Innovation, vol. 10, no. 2, Article 2008, 2026. [2]    M. E. Haque, M. T. H. Saykat, A. Al-Imran, A. H. Siam, J. Uddin, and D. Ghose, “An attention enhanced CNN ensemble for interpretable and accurate cotton leaf disease classification,” Scientific Reports, vol. 16, Article 4476, 2026. [3]   P. Johri, S. Kim, K. Dixit, P. Sharma, B. Kakkar, Y. Kumar, J. Shafi, and M. F. Ijaz, “Advanced deep transfer learning techniques for efficient detection of cotton plant diseases,” Frontiers in Plant Science, vol. 15, Article 1441117, 2024. [4]   S. Chinnadurai and S. Selvakumar, “Image-based cotton leaf disease diagnosis using YOLO and Faster R-CNN techniques,” Scientific Reports, vol. 15, Article 45785, 2025. [5]  F. Hu et al., “Investigation of an efficient multi-class cotton leaf disease detection algorithm that leverages YOLOv11,” Sensors, vol. 25, Article 4432, 2025. [6]  M. Naeem et al., “Automatic prediction of cotton leaf’s diseases using deep learning techniques,” Scientific Reports, 2026. [7]   D. Ram, A. Avadhanam, M. Gyanchandani, A. Rasool, and N. Khare, “Federated deep learning for cotton leaf disease detection in Indian agricultural environments using EfficientNet B0 and FedSCAFFOLD,” Discover Artificial Intelligence, vol. 6, Article 614, 2026. [8]   M. S. Memon, P. Kumar, and R. Iqbal, “Meta deep learn leaf disease identification model for cotton crop,” Computers, vol. 11, Article 102, 2022. [9]  R. F. Caldeira, W. E. Santiago, and B. M. Teruel, “Identification of cotton leaf lesions using deep learning techniques,” Sensors, vol. 21, no. 9, Article 3169, 2021. [10]    D. Zhu, Q. Feng, J. Zhang, and W. Yang, “Cotton disease identification method based on pruning,” Frontiers in Plant Science, vol. 13, Article 1038791, 2022.