Very cool, I look forward to it!
As a visual person, I'm really curious what additional parameters you're going to look at... Like which images were AI generated, or was the image the theme of the post (like an artist's creative process) or was the image hand-drawn vs. photo, etc.
I can't even imagine what work goes into extracting all this data, so please don't feel obligated to go that far... I'm just throwing ideas.
I won't be able to go down that level of depth with the tech (and scale) I have st my personal disposal, but there might be ways to see if of image captions are mentioning a source or if they're tagged as AI generated by the user when giving credit. There are also going to be a lot of image strings in the text that are things like authors signatures, or dividers, that aren't really "pictures" thay add value to the post.
There's a lot of variables. I'll only know what is capable as I spend a few more hours each week refining my methodology.
Hopefully people will continue finding it interesting or useful.