This thread is specifically devoted to discussion of any pre-existing art style resources that could be used in the Artistic Style Taxonomy project we're trying to get going in this group.
The key way to get things done in the modern world is to leverage off of pre-existing resources. So if it's software, you want to be looking for good open source libraries you can build on top of. What you do not want to do is re-invent the wheel.
This axiom also applies to the Artistic Style Taxonomy database. We want to identify and take advantage of as many pre-existing art style resources we can get our grubby little hands on. If someone has already done the work for us, let's take advantage of that and build on top of that good work.
I'm putting together some ideas for resources we could take advantage of. And i will get around shortly to doing some sub posts in this discussion to talk about them.
But i'm sure some of you might be aware of particular resources, databases of images, etc directly related to your artwork. Or styles of artwork you particularly like.
So feel free to point them out to everyone in this discussion thread.
Replies
Here's another Kaggle dataset i stumbled upon while researching something else that i'm going to notate here. Since it would probably be fun for someone to muck with.
Here's a link to the CelebFaces Attributes (CelebA) Dataset.
The CelebFaces Attributes Dataset (CelebA) is a large-scale face attributes dataset with more than 200K celebrity images, each with 40 attribute annotations. The images in this dataset cover large pose variations and background clutter. CelebA has large diversities, large quantities, and rich annotations, including
Another tack to 'pre-existing resource' is to custom curate our own dataset, but take advantage of pre-existing resources like Bing image search. I mention Bing because they apparently have some specific service you can use to auto-download image searches as a folder of images. You are throttled so that AI-Bots and other autonomous systems can't use it for nefarious purposes, but for out purposes of building up a dataset of high quality targeted artistic style images it seems quite useful.
So, some of us could be using this to put together a curated set of images for different components of the Artistic Style Taxonomy. So this is your chance to make sure artistic styles that are important to you are included in the Taxonomy.
Some comments that i have received from different artist about this go along the general lines of, 'boy these academic art style datasets are very stilted' or 'boy they sure are biased', or 'do not reflection of the actual artwork i personally am interested in', etc. Building our own dataset is a way to try and avoid these problems.
Maybe our more targeted curated datasets are ultimately included along with other academic or project datasets into some larger master 'dump it all in here' dataset that might be useful for some tasks. For other tasks, speciallized style specific datasets might make more sense.
I can put together some more info about how to use Bing for what i'm talking about. I'll add that as an additional post in this thread when i get around to it. I'm sure you can figure it out online in about 2 minutes if you are dying to get started yourself, so please feel free to dive in.
One pre-existing art style web resource is WikiArt.
WikiArt's stated goal is to make the world's art accessible to anyone. It features over 250,000 artworks by over 3000 artists.
So is there an api to access it as a database, or download, or? Someone feel free to look into this. I will get around to it at some point otherwise.
There are some other downloadable databases that claim to use images from WikiArt.
One example would be the Kaggle Painter by Numbers competition.
Painting-91 is a novel large scale dataset of digital paintings. It is composed of paintings from 91 different painters.
The Paintings Dataset was put together by Andrew Zisserman's group at Oxford.
'WikiProject sum of all paintings' project is a WikiProject to have a Wikidata item for every notable painting.
A dataset of 14,912 landscape paintings was used in a study (and associated paper) called 'Dissecting Landscape Art History with Information Theory'. The data is available for download here.
Pandora is a database of more than 7700 images from 12 art movements.
So there's a quick list of a few pre-existing resources to think about.
What's good about them? What's bad about them?
If anyone feels that one or more of these are datasets we should be working with, please feel free to figure out how to access the data and then provide that explanation here so that others can access it as well. And at that point we can get a little bit more organized about how we're going to organize it for use in out Artistic Styles Taxonomy.
John, I just had a quick look on the web and found this; https://www.programmableweb.com/category/art/api
112 Art ApIs, WikiArt is in there plus lots of other databases, hopefully thats a start.
Cool. Thanks for pointing that out.
One thing to be aware of is that sometimes the museum databases imply you have access to the images, but really what you are getting is just the meta-data, and then to get actual image data they want you to purchase individual art images.
The MOMA dataset is like that, you just get the meta data. Which could still be potentially useful in some way. But you'd probably do better just getting on Bing and building up a MOMA image set over a few days of targeted searches if you wanted images from that kind of thing.
I guess a nefarious individual or AI Bot could use the meta data to run image searches on the web and grab images from the return results of the search off of the meta data for individual art pieces in the collection.