Retracted: A Comparison of Two Transformers in the Study of Plant Disease Classification
V. R. N. Murthy Teki, R. Anandha Ragaven, N V Sai Manoj, V Vipul, S Sarath
Abstract
This paper, have implement two transformer models, namely Vision Transformer (ViT) and Shifted WIndow Transformer (Swin) for identifying paddy leaf diseases at an early stage, using thermal images. Thermal images provide detailed and precise information when compared to normal images, which helps in the early detection of leaf diseases, even before the spots appear on the leaves. The aim of this paper is to prove that both Vision as well as Swin transformer works well with vision (image) tasks, such as image classification. Here, in our paper we were able to identify and classify paddy leaves diseases using ViT and swin transformer with an accuracy of 94% and 98% respectively. Swin transformer works better when it comes to semantic segmentation, where one needs to classify each pixel to it’s respective category. It does so by using the concept of Windows, about which we will talk in detail in the coming sections.