Digit Recognizer

1 0
  • 0 Collaborators

A Digit Recognizer Demo Application Completely Built in Ionic Framework, with the help of Keras Trained and Developed Model, Changed to TensorflowJS model with the use of Tensorflow JS Library. This project shows the demo of Basic Working of TensorflowJS and Ionic Framework as a Web App Development Framework. ...learn more

Project status: Published/In Market

Mobile, Artificial Intelligence

Code Samples [1]Links [1]

Overview / Usage

This Project is for demoing purpose but shows the immense opportunities for vaious domains. Many production ready applications could be built using the TensorflowJS technology which would have Offline running capabilities.
It shows the perfect project structure of Ionic Framework and it's usage for web app development framework.
This project, if extended, and could have a bit more work, could be used for better OCR Development. Teachers teaching online with the help of Paint or Stylus Screen, could provide students lecture notes without doing much extra work.
Seminars and Webinars will have plane text recorded if this could be used.

Methodology / Approach

The Project was developed in two different phases.
Initially the model was trained using the MNIST Dataset and Keras Framework. All the Model Building, Validation, Training and Testing was done in this phase. Model Accuracy was 99.3% and hence was acceptable. The model was then converted into a TensorflowJS model and was saved in the Ionic App project as an Offline resource for the app to use.

The second phase was app building. The app has a on screen drawing capability in a Box. This box is then broken into a 28x28 pixel parts. This frame is then feeded into the Model with the help of library TensorflowJS, and the predictions are then recorded and projected in the Bar Graph.

Technologies Used

3 Main Technologies used were:-

  1. Ionic Framework
  2. Keras Framework
  3. TensorflowJS Library

Repository

https://github.com/Rish1997/Digit-Recognizer

Comments (0)