youssofkarhani / number-recogintion

A number recognition program, where the user trains the neural network with numbers that he draws on the canvas and then asks the model to guess the new input.

Number Recogintion

Overview
This project is a neural network based image/number recognition system built using HTML, CSS, and JavaScript. The user has an 8x10 grid to draw a number, which is then used to train the neural network. The neural network is implemented manually, with a focus on understanding the underlying matrix operations.

Usage
Open the project in github pages or https://youssofkarhani.github.io/number-recogintion/. Draw a number on the 8x10 grid using the mouse. Press the "Train" button to feed the drawn number into the neural network for training. Repeat this process for multiple numbers to improve the accuracy of the model. Once the model is trained, draw a number on the grid and press the "Recognize" button. The model will display the recognized number and the percentage of confidence in the console.

Note
The model's accuracy may vary depending on the number and amount of training data provided. Feel free to experiment with different training sets and tweak the neural network's parameters to improve performance.