Implementing additional features extraction using timm.

[AgenoDrei]

I tested using the Pytorch Image Models (timm) library to extract RETFound and Dino v2 features for my own research project. The timm library makes this pretty simple by giving a unified format for models and their configuration.

Would you be interested in dino v2 features using this library for your own project? I could start working on a fork in a few weeks.

[v-iashin]

Hi. We support extraction with timm models.

https://v-iashin.github.io/video_features/models/timm/

Do you mean that something is not working?

[AgenoDrei]

Hey, I did miss that in the documentation. Nevermind and thanks for providing the great library! :)