Patrick Labatut

> Windows 10 The code was only tested under a Linux environment (and training and evaluation do require such environment). Some other users have suggested using WSL, but this is...

Closing as answered, thanks for your interest!

> the part for instance retrieval is not public The k-NN evaluation code does provide functionality close to that though (with a k-NN classifier). To complement @MarcSzafraniec's answer with a...

Closing and keeping track as part of #95.

No, we have not (yet), but we might consider alternative architectures for the distilled models.

Closing to keep track of similar asks in #166 instead

@Suhail To generate features from the pretrained backbones, just use a transform similar to the standard one used for evaluating on image classification with the typical ImageNet normalization mean and...

Hi and thanks for your interest in using DINOv2! No, all the visual tasks on which we had evaluation results are listed in paper and we haven't tried (yet) OCR...

Closing as this seems resolved (and using #53 to keep track of documentation needs on feature extraction).

Thanks for your interest! That request is related to #6 and we are monitoring feedback to possibly provide more pointers or helpers to reproduce the results, including depth estimation and...