Painfree LaTeX with Optical Character Recognition and Machine Learning

Recent years have seen an increasing interest in harnessing advancements machine learning (ML) and optical character recognition (OCR) to convert physical and handwritten documents into digital versions. The increasing adoption of digital documents in academia, however, has provided a new layer of complexity to automatic digitization of physical documents. Compared to typical texts written in natural language, academic papers often contain many elements like equations that are notably tougher to recognize with many current OCR techniques. Furthermore, creating digital documents from scratch tends to be difficult too: while typesetting systems like LaTeX help streamline equation formatting and document generation, the process of typesetting complex equations is a arduous and error-prone process. Ultimately, academics aspire to find a solution where they can get the simplicity of generating documents by handwriting while still obtaining the convenience of using digital documents. In this paper, we aim to explore various techniques for converting images of mathematical equations into LaTeX source code.