during my master thesis I'm working on / towards an OCR for math. For testing purposes I need a massive set of mathematical equations of various formats (font family, font size, used equation formatter etc.) I did some googling and could only find corpora for equations as text, rather than as images.
Since an OCR should render images and retrieve the text, those corpora don't help me. I thought about recreating the images, e.g. if the equation is given as TeX, but that would mean that I get pretty similar looking equations.