Handwriting Classification

Role: Founder & Sole Developer
Duration: September – November 2021
Website

Purpose
Handwritten digits are currently labeled as distinct labels, yet there are cases in which digits can be seen in two different ways. I wanted to create a site where users can vote on what digit they see, creating a probability distribution to be used as training data for uncertain or fuzzy predictions.

Tasks

  1. Data Collection

I created a website where users can vote on a total of 100 questions. Users are not required to vote for all 100, and can submit as many answers as they want.

  1. Use machine learning to build a classification engine based on collected data
  2. Data analysis – analyze ambiguous digits to see the accuracy

Improvements
UI improvement on voting website – let users fill out more than 10 questions before submission

More To Explore

EM Touch App

EMTouch is an app that trains users to identify the emotions behind common facial expressions, with the goal of facilitating communication between autistic and neurotypical individuals.

Astronomy App

App that informers users to special astronomical phenomena and enables astronomers to share their data and findings through an online platform