https://github.com/maneprajakta/digit_recognition_web_app

A Hand Written Digit Recognition app trained on the MNIST dataset of Keras using the CNN model. skills used are Tensorflow, HTML,CSS,javascript.
https://github.com/maneprajakta/digit_recognition_web_app

convolutional-neural-networks digit-recognition machile-learning mnist-dataset webapp

Last synced: about 2 months ago
JSON representation

A Hand Written Digit Recognition app trained on the MNIST dataset of Keras using the CNN model. skills used are Tensorflow, HTML,CSS,javascript.

• Host: GitHub
• URL: https://github.com/maneprajakta/digit_recognition_web_app
• Owner: maneprajakta
• License: mit
• Created: 2020-06-21T07:33:30.000Z (almost 5 years ago)
• Default Branch: master
• Last Pushed: 2020-06-21T17:10:29.000Z (almost 5 years ago)
• Last Synced: 2025-02-27T18:06:58.359Z (3 months ago)
• Topics: convolutional-neural-networks, digit-recognition, machile-learning, mnist-dataset, webapp
• Language: Jupyter Notebook
• Homepage: https://maneprajakta.github.io/Digit_Recognition_Web_App/
• Size: 1.42 MB
• Stars: 22
• Watchers: 2
• Forks: 2
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        

# Digit_Recognition_Web_App

 link : https://maneprajakta.github.io/Digit_Recognition_Web_App/




Structure of App

 keras - > Tensorflow.js ->(html + css + javascript)->github pages

  
Hello World of Object Recognition!

 
Aim:
 To make a convolution neural network to recognise handwritten digits by training the model on MNIST dataset available in keras.

 


 
MNIST DATASET:
The training dataset contain 60000 images and testing contain 10000 images .Each image is 28x28 pixel and grey scale.

  


 
CNN MODEL OVERVIEW:

 
⚈ It is a 17 layer model with Conv2D,MaxPooling2D,BatchNormalization,Dense,Flatten and Dropout layer combination.

 
⚈ Input layer has 32 neuron and output layer has 10 neurons as 10 different clases exsist.

 
⚈ 30 epochs are used.

 
⚈ Categorical_loss is loss function and adam is used for optimization.

 
⚈ Model gives 99.15% accuracy.

For Deployment:
Save model using tensorflowjs converters as json file and weight as .h5 file.Use Tensorflow.js to load model and predict in javascript file