Math Number Checker

Building our AI Model Layers - Code
Training and Validation Plots
Training and Validation Datasets - Code
Creating our Training and Validation Plots - Code
Our Website

Inspiration

This project was created simply to make the lives of middle-school teachers easier. We wanted to create a model as a basis that could be built upon in the future to create an automatic grading AI that would alleviate some of the busy work that school-teachers have to go through. The grading process isn’t difficult but it is tedious and takes away the time that teachers have to really focus on the students. Ultimately, our goal is to enhance the teaching experience, improve efficiency, and enable teachers to foster a more engaging and impactful learning environment for their students.

What it does

This AI model is able to take in an image and recognize the numbers that are displayed on it.

How we built it

We built the backend model through Google Collab. We have a set of over 10000 images all of which have different variations of commonly seen numbers and variables. By creating a .2 variation split we were able to train the model to recognize the different formations. We also built the frontend website within google apps-script: coding it with a mix of css, javascript, and html.

Challenges we ran into

The single biggest challenge that occurred was connecting the front and back-ends of the code. It was extremely difficult to link up the different languages, libraries, and downloads together to make one fully functioning app.

Accomplishments that we're proud of

We are extremely proud of the accuracy of the model and its ability to recognize the different letters and variables after an ungodly number of edits to the code. We are also pleased with the aesthetic of the website after learning html and css in the little time that we had.

What we learned

We learned the basics of 2 new languages and a large amount of information related to AI based coding. This includes the different libraries and what they do, how HTML and CSS specifically have the ability to link up and generate the objects. By understanding these concepts and techniques, we have expanded our repertoire to create AI-based solutions and leverage the power of data-driven decision making.

What's next for Math Number Checker

We plan to continue this project to complete its original purpose of grading papers on its own. After it receives a 100% accuracy in recognition it will be able to check the answers that the teachers have provided and understand if they are the same. Furthermore, we hope that it will be able to recognize errors that may have occurred within the work itself to save them time and effort in the future.