ResViT: A Hybrid Vision Transformer Model for Onychomycosis Detection

Onychomycosis is a very common fungal nail disease. It has no life risk but can be sickening. If it is not diagnosed at an early stage, it can be severe. AI can play a vital role in detecting Onychomycosis precisely and faster. We used a hybrid network, ResViT, consisting of a pre-trained ResNet-50 model for feature extraction and a Vision Transformer (ViT) for classification in our paper to detect Onychomycosis. As we got a small dataset consisting of 1355 diseased data, we could not use conventional ViT. We got 99.77 percent accuracy using our hybrid ResViT model, which was identical to the pre-trained ResNet-50 model. But for conventional ViT, our accuracy decreased to 96.99 percent while time per epoch increased. Overall, our proposed hybrid ResViT performed better than any state-of-the art techniques.