Very exciting

!

I love this multidisciplinary application of ML! Two questions (sorry if you already addressed either of these!): (1) Given the small size of the patches, do you expect you will have to make any changes to your model or data mapping to compare paintings of different subjects? (In other words, how generalizable do you think your model is?) (2) Do you think you could use some form of this model to classify some sample of a painting as representative of a painter’s late style or as representative of a painter’s early style?

Thank you!!

what do you consider to be the optimal size of a training set and how to achieve that with El Greco

Love the thought dedicated to the ethical and aesthetic evaluation of replicated art objects. Some of this work seems like a vital corrective to the practice of extracting art/artifacts from their local contexts, often at great cost to the people who live there or the art itself. But is there anything we lose just by viewing art outside its intended context, even a perfect replica?

How about for artists who don't use brush? If a painter uses a different tool to apply paint, does that change methodology at all?

(sorry if this was addressed, my wifi briefly cut out)

Wow. That's very compelling. I agree.

So interesting. Thank you!

Just out of curiosity, do brushes have their own signature?

What an interesting set of questions this project raises. Thanks!

I think what is particularly interesting about this project is working/training with customized datasets because most image recognition work done in art history is done around digital surrogates and photographs- e.g. photo archives

you love to see it!

interesting way to also overcome the constraints of small datasets - see the work of Rick Johnson/Cornell - watermarks- felt that the dataset was too small for machine learning

Thanks everyone!!

Thank you!