Handwritten Mathematical Equation Recognition and Solver

Abstract

Our contribution to the field of Handwritten Math- ematical Equation Recognition is the development of an end-to- end pipeline that combines character recognition and equation solving. Both areas have been extensively worked on individually, hence we aim to combine both pieces to form a complete user application. Recognition will be performed by a pipeline consisting of Image Cleaning, Segmentation, and Recognition. A shallow Convolutional Neural Network performs recognition and the SymPy math engine solves the recognized equation. We have also included a feedback mechanism to correct anyfalsely classified symbols. The proposed system is tested on the CROHME dataset, and the model accuracy is tested along with user interface testing. To demonstrate the final system, we have also created a Graphical User Interface (GUI) that provides the user with options to handwrite equations, upload images of equations, interact with graphs and provide feedback on incorrect equations.