Cotton is a vital cash crop globally, and diseases can cause signifi-cant damage, leading to reduced yields and quality. Early detection enables farmers to take necessary measures to curb the spread of diseases, enhancing crop quality and yields. Typically, cotton diseases have been identified by ex-perts via visual inspection, which is costly, time-consuming, and not always precise. However, recent advances in deep learning and computer vision pro-vide new opportunities for automated disease detection. This research paper presents a model for detecting cotton crop diseases using the SWIN transformer architecture and Attention-Based CNN model. The proposed system was trained on the large and small datasets of images of healthy and diseased cotton crops, covering three different diseases named Bacterial Blight, Curl Virus, and Fusarium Wilt. The Swin Transformer architecture was chosen for larger da-tasets and Attention-based CNN for smaller datasets due to their superior per-formances compared to other state-of-the-art deep learning models as it achieved better accuracies. The ultimate goal is to provide farmers with an ef-fective tool for early detection and prevention of cotton crop diseases, improv-ing crop yield and reducing economic losses. The success of this project could lead to further advancements in deep learning solutions in agriculture, ultimate-ly enhancing global food security.