Cucurbit Leaf Disease Classification using Compact Vision Transformer and Cubic Support Vector Machine with Metaheuristic Optimization

Monisha, R.; Tamilselvan, K. S.; Karthiba, L.

doi:10.24200/sci.2026.67174.10466

Cucurbit Leaf Disease Classification using Compact Vision Transformer and Cubic Support Vector Machine with Metaheuristic Optimization

Articles in Press

Document Type : Research Article

Authors

R. Monisha ¹

K. S. Tamilselvan ¹

L. Karthiba ²

¹ Department of Biomedical Engineering, KPR Institute of Engineering and Technology, Coimbatore, Tamil Nadu, India.

² Department of Plant Pathology, Tamil Nadu Agricultural University, Coimbatore, Tamil Nadu, India.

10.24200/sci.2026.67174.10466

Abstract

Cucurbits, a widely cultivated crop from the Cucurbitaceae family, account for about 6% of global vegetable production. Their yield, however, is susceptible to disease and pest attacks, contributing to food insecurity. Timely detection of these diseases is crucial to prune agricultural yield losses. This paper proposes a hybrid compact vision transformer (CVT) and cubic support vector model (CSVM) that employs Harris Hawks Optimization (HHO) for cucurbit leaf disease classification. Initially, image preprocessing is carried out using histogram equalization and GrabCut algorithm to effectively handle images with dispersed backgrounds. The preprocessed output is fed into the proposed hybrid model, where the CVT hyperparameters are optimized using the HHO algorithm. Finally, the features are extracted through CVT and classified using CSVM. The efficiency of the proposed model is highlighted by comparing it to various standard machine learning classifiers. Experimental study with the cucurbit dataset reveals that the proposed method attained 96.96% accuracy, contributing to sustainable crop production.

Keywords

Compact vision transformers

Cucurbit disease detection

Deep learning

Harris Hawks optimization

Hyperparameters

Machine learning

Subjects