Vision Transformers (ViT)

Token-to-Patch Reconstruction

Reverse process in generative tasks where tokens are converted back into image patches to reconstruct the original image.

← Indietro