Lightweight Vision Transformer for Efficient Influenza Virus Subtype Classification via Genomic Image Processing

This article presents an innovative approach to classify influenza virus subtypes A, B, C, and D. Accurate classification is crucial for public health since these variants show varying degrees of severity and transmission. Our suggested approach uses methods like the Grammian Angular Summation Field and Grammian Angular Difference Field and their modified versions to convert DNA sequences into visual patterns. These patterns are subsequently processed using low-parameter models like MobileNetV2 and Vision Transformers to identify subtypes accurately. This approach obtains a classification accuracy of 98.34% and is less labor-intensive and efficient than existing methods.