AI-Powered LaTeX Generation from Handwritten Mathematical Equations

This document discusses the creation of a real-time web application that identifies handwritten mathematical equations and transforms them into syntactically accurate LaTeX code employing a pre-trained deep learning model.The model used in this system was trained on a collection of 100,000 authentic handwritten mathematical equations, enabling it to generalize effectively across many different handwriting styles.The application employs ONNX Runtime to load and run the trained model, while Streamlit is utilized to develop an interactive frontend accessible via browsers for image uploads, LaTeX rendering, and result exporting.The system prioritizes usability, modularity, and reliability within educational settings.It doesn't carry out model training; instead, it emphasizes the smooth integration of current AI features into a user-friendly interface that includes preprocessing, postprocessing, LaTeX generation, copy-to-clipboard, and export options.Testing through experiments verifies that the application can reliably generate precise LaTeX from handwritten equations that are single-line.