I understand the UI is not to everyone's taste, but I think the idea of using machine learning to create an engaging "window" into large sets of public-domain images is a good one. It would be interesting to read more about the process used to tag and organize the content for this project.

There are lots of other repositories of high-quality free content (museum archives, Flickr, Wikimedia Commons, etc) that could benefit from this kind of machine-aided presentation.