Visual Digits :TRAINABLE K-NN based ML thinking model
Link to open source: https://github.com/bhuvan-rai/visualdigits
Link to Live Project: https://bhuvan-rai.github.io/visualdigits/
#Visual Digits
A webapp made in vscode that guesses cursorwritten digits from 0-9 using a user-trainable K-NN model
The app uses a simple manually trained dataset anyone can update locally by training their K-NN model , of which new training data is stored in localstorage of the browser itself !!
## Features
- Draw digits with your cursor
- Guess digits from 0 to 9
- Manually train the webapp by approving the correct digit
- Saves training data locally in the browser's localstorage
## Training Your Dataset- after drawing and guessing , you can click one of the digits from 0 to 9 below the "train as" header
- on clicking , the guessed drawing is stored under that digit as sample
- the more data the better
## How It Works
- Your drawing is converted into a 28×28 grid of pixel values. When you click Guess, the app compares it to every saved example using Euclidean distance and picks the closest match (k-NN with k=3). The pre-trained dataset (217 samples) is embedded in the js file from my python project. Any new drawings you train are saved to your browser's localStorage and persist across page loads.
