Vision Transformers for Detection

Token-to-Token ViT

Variant introducing a progressive transition between tokens with resizing and recombination to preserve local structural information.

← Back