MathX Recognizer

18K Shake It • [Kuck Swee Nien - Lim Wei Xin - Nah Wan Jun - Law Shiao Yin]
15 Jun 2022

MathX Recognizer
WIE2003 Group project using Machine Learning and Data Science Analysis Process on mathematical expression image data.

What Do MathX Recognizer Solve?

This presentation reports the Group Project of WIE2003 Introduction to Data Science course guided by Dr Salimah.

This Shiny app is built entirely in R, with API connected to our TensorFlow model using HTTR library, with purpose to predict mathematical expression.

Steps in predicting Maths Symbol:

Story

Data Science Process

Asking Questions
  • Can computer predict mathematical symbols from images?
  • How computer differentiate symbols through pixels?
  • Who are the stakeholders?
Finding, Getting & Cleaning Data
  • The dataset is from Xai Nang called Handwritten Math Symbols Dataset
  • This dataset contains over 100000 image of pixel 45*45
  • It contains 17 expressions [+ - x / ( ) 0 1 2 3 4 5 6 7 8 9 =]
  • Image of 45*45 is vectorized into vector of 2025 and stored in matrix

Experience Creating MathX Recognizer

This is a prototype for Mathematics Expression Recognition App!

Due to memory, computation processing power limitation on my computer and also shinyapp.io restriction:

  • The analysis use only 13% of the entire data provided.
  • The prediction model may not be accurate due to this constraint.

It has been a challenging though fruitful experience to complete this project. We have learnt to use TensorFlow, API, HTTR, Shiny, and RPubs!

MathX Recognizer can be accessed at Shinyapp.io & GitHub

Inside MathX Recognizer

  • The app starts with “Main” tab, that prompt users to upload image of mathematical expression.
    Main Tab
  • “Data Views” and “Data Analysis” tabs shows descriptive and exploratory analysis. Main Tab Main Tab Main Tab Main Tab Main Tab

Key Takeaways

Answering: Who are the stakeholders?

Of course, students!

Keys

1) MathX can recognize mathematical expression from image

2) MathX have the potential to learn more - and eventually become the greatest calculator or mathematician

3) MathX with its easy to use attribute, will become must have apps for students

4) Business model of ads and subscription