Font Recognition Reimagined with FasterViT-2

Collabora extended the AdobeVFR dataset and trained a FasterViT-2 font recognition model on millions of samples. The result is a state-of-the-art model for fine-grained font identification that can also be used for downstream tasks.