Figure Search Engine I

I’ve finished mining ~ 1M figures from a venerable preprint server. It was a great learning experience, and I’ll write more shortly on the findings. I’m about to release the data, and to that end put together a boilerplate MEAN app for serving up the details using AWS on github. Each figure from > 100k papers is modelled as a Gaussian mixture, and through the web app you’ll have access to figure images, as well as CSV spreadsheets with details on the model for a given figure. More soon …