MirFlikr Dino V2 embeddings

This is a dataset created from the 2010 MIRFLICKR-1M dataset (https://press.liacs.nl/mirflickr/) The MIRFLICKR-1M dataset contains 1 million Flickr images under the Creative Commons license. In this repository we store a set of embeddings of this dataset created using the Dino V2 encoder (https://github.com/facebookresearch/dinov2). Each of the embeddings is 384 dimensions and has been L2 normed. The same data is supplied in both csv and h5 formats.