/digdet

A realtime digit OCR on the browser using Machine Learning

Primary LanguageJupyter NotebookMIT LicenseMIT

A realtime digit OCR on the browser using Machine Learning

Demo

You can view the demo on https://digdet.herokuapp.com

Table of Contents

Installation

Download or clone the repository as follows:

$ git clone git@github.com:achillesrasquinha/digdet.git

Install necessary dependencies:

$ make install

Usage

$ make all

Your app should now be up and running on http://localhost:5000.

API

Send in a base-64 encoded image string HTTP POST (with image as the parameter) request to https://digdet.herokuapp.com/api/detect. You'd then recieve a sample response as follows:

{
    "id": "<unique_response_id>",
    "version": "<api_version>",
    "data":
    [
      {
        "digit": "<recognized_digit>",
        "rect":
        {
          "x": "<x_coordinate>",
          "y": "<y_coordinate>",
          "width": "<contour_width>",
          "height": "<contour_height>"
        }
      }
    ]
}

Dependencies

  • Python 2.7 and more or 3.5 and more
  • Node.js
  • SASS

License

This repository has been released under the MIT License.