First project of 2025: Vision Transformer Explorer
I built a web app to interactively explore the self-attention maps produced by ViTs. This explains what the model is focusing on when making predictions, and provides insights into its inner workings! 🤯