train on coustom dataset #44
Description
Hi,
Thank you very much for the great work you’ve done. I have a question: can I use the output of preds = model.match(im1_path, im2_path) from RoMa v2 to obtain labels for my own images, and then fine-tune Tiny RoMa v1 that you published in your previous paper?