1Ph.D Research Scholar, Department of Computer Science, PKR Arts College for Women, Gobi, Bharathiyar University, Coimbatore, Tamil Nadu, India
2Associate Professor, Dean and Supervisor, Department of Computer Science, PKR Arts College for Women, Gobi, Bharathiyar University, Coimbatore, Tamil Nadu, India
Background: The integration of artificial intelligence with agricultural science has significantly advanced precision farming by enabling automated monitoring and diagnosis of crop health. In this study, a hybrid deep learning and statistical-based framework is proposed for efficient tomato leaf disease detection and classification. The system begins by pre-processing the input leaf images using a Lagrange Conditional Generative Adversarial Network (LCGAN), which enhances image contrast and reduces background noise for improved segmentation accuracy. The Weight Hippopotamus Optimization Multi-Threshold Segmentation (WHOAMTS) technique is then employed to effectively separate diseased areas from healthy regions.
Methodology: Subsequently, a hybrid Statistical Feature–Based Inception-V3 model is utilized for feature extraction, combining deep convolutional representations with handcrafted statistical features, including GLCM-based texture descriptors, colour moments, and geometric shape parameters. This integration ensures comprehensive capture of both high-level semantic information and fine-grained spatial details. The extracted hybrid features are further refined using the Improved Grey Wolf Optimization (IGWO) algorithm to select the most discriminative feature subset, thereby enhancing classification performance while reducing redundancy.
Classification: Finally, the optimized features are classified using an Optimized Multi-Scale Attention Multi-Axis Vision Transformer (OMAMViT), where the transformer's attention weights and hyperparameters are fine-tuned using the Animal Migration Optimization (AMO) algorithm. The proposed hybrid approach leverages statistical–deep feature fusion, intelligent feature selection, and metaheuristic transformer optimization to achieve superior accuracy, faster convergence, and robust generalization across multiple tomato leaf disease categories under diverse imaging conditions.
Keywords: Tomato leaf disease detection, Lagrange Conditional GAN (LCGAN), Weight Hippopotamus Optimization Multi-Threshold Segmentation (WHOAMTS), Improved Grey Wolf Optimization (IGWO), Optimized Multi-Scale Attention Multi-Axis Vision Transformer (OMAMViT), Animal Migration Optimization (AMO).
How to cite this article: Tamil Ilakkiya NS, Gomathi PM. Tomato Leaf Disease Detection Using IGWO Feature Selection and AMO-Optimized Multi-Scale Attention Multi-Axis Vision Transformer. Int J Drug Deliv Technol. 2026;16(13s): 950-974. DOI: 10.25258/ijddt.16.13s.108.
Source of support: Nil.
Conflict of interest: None