Here, in this study, we have described an image classification and sorting technique with vision transformers (ViTs). Moreover, we elaborated on how the dataset was organized, the categorization of images for each class, and the procedure to load and preprocess the data. The method involves splitting images into patches, encoding them, and then using a transformer encoder neural network to do the classification tasks. The metrics were used to report the model performance with accuracy metrics and loss curves for the specific goals, mentioning the difficulties of training the model with small dataset and low epochs. This work improves the knowledge of ViTs-based image classification and sorting, both accuracy and robustness, using deep learning, where it establishes a baseline for further work in this area.