<p>Cotton is a crucial role in the world textile trade, which is threatened by environmental factors and climate change, Making yield prediction necessary for both sustainability and economic stability. Conventionally, remote sensing-assisted crop yield estimation is usually done using Machine learning methods. However, conventional techniques using machine learning may face problems when using yield regression functions in farming. Hence, a new deep learning-aided cotton yield prediction framework is implemented using remote sensing images. At first, the required remote sensing images are aggregated from available data sources. Then, 3D Yolo-Transformer Unet++ (3D YTUnet++) is introduced and the 3D convolution in this model helps to extract hidden features and segment anomalies. The segmented images are then processed through a Vision Transformer (ViT)-aided Adaptive ShufflenetV2 (ViT-ASNetv2) for accurate yield prediction. Here, the Improved Hermit Crab Optimizer (IHCO) is employed for optimizing the parameters of the ShufflenetV2 technique. The performance evaluation represents with integration of these sophisticated techniques provides a more accurate and reliable model for predicting cotton yields from remote sensing data. The accuracy score achieved by the designed model in cotton yield prediction is 96.4%.</p>