README

This project involves building a Vision Transformer (ViT) from scratch and training it on the MiniPlaces dataset to explore the capabilities of transformer-based architectures in image classification. Additionally, the project includes constructing a Semantic Segmentation model using a ViT encoder, providing a deeper understanding of how ViTs can be applied to pixel-level prediction tasks. This comprehensive approach allows for a thorough evaluation of ViTs in both image classification and semantic segmentation.