Classification Optimization of Aedes albopictus and Culex quinquefasciatus Mosquito Larvae Using Vision Transformer Method

Abstract

Mosquito-transmitted diseases like Dengue Hemorrhagic Fever and Filariasis pose serious health threats throughout tropical regions, particularly in Indonesia. Quick and accurate identification of mosquito larvae plays a crucial role in disease prevention, especially for Aedes albopictus and Culex quinquefasciatus species that act as main disease carriers. Manual identification methods using microscopes or visual guides often struggle with time constraints, accuracy issues, and dependence on trained specialists. Our research focuses on improving the classification of Aedes albopictus and Culex quinquefasciatus mosquito larvae using Vision Transformer (ViT) technology, a deep learning method that has shown strong results in image recognition tasks. We applied the Vision Transformer model to classify mosquito larvae from microscopic field images. The study also tested how different factors impact model performance, such as image clarity, lighting conditions, and image resolution. Our findings show that using Vision Transformer in classification systems produced excellent results, achieving 98.00% accuracy in recall, precision, and F1-score measurements. The research reveals that Vision Transformer methods deliver better accuracy than traditional approaches like Convolutional Neural Networks and can be adapted into working systems for technology and healthcare sectors.