GCP Vision API Document Recognition (Demo)

Simple interaction with GCP Vision API to provide a Document Recognition Service.

The aim of this project is to simply display some of the capabilities of the GCP Vision API

Currently supported documents:

• Victorian Driver's Licence

Additional documents can be supported by contributing to the templates available in the code base, which are list of tuples, describing the position of each bounding vertex of a textual component of the document, VIC_LICENCE_NUMBER = [(77.34, 21.27), (95.74, 21.27), (96.06, 27.63), (78.72, 27.13)] The above template translates to vertices (top-left, top-right, bottom-right, bottom-left) in which the licence number is located on a Victorian Driver Licence.

Install Dependencies

TODO

Set GCP Client Credentials

Be sure to create an account with GCP and obtain your credentials in JSON format and add them to your Terminal session: export GOOGLE_APPLICATION_CREDENTIALS=/path/to/credentials.json

Run the Application

python3 api.py

Upload Image to Recognize

Request:

POST /documentRecognitions HTTP/1.1
Host: 127.0.0.1:5000
Cache-Control: no-cache
Content-Type: multipart/form-data; boundary=----WebKitFormBoundary7MA4YWxkTrZu0gW

------WebKitFormBoundary7MA4YWxkTrZu0gW
Content-Disposition: form-data; name="image"; filename="vic_licence.jpg"
Content-Type: image/jpeg


------WebKitFormBoundary7MA4YWxkTrZu0gW--

Response:

{
    "data": [
        {
            "Key": "VIC_LICENCE_NUMBER",
            "Value": "987654321"
        },
        {
            "Key": "VIC_LICENCE_EXPIRY",
            "Value": "20-05-2019"
        }
    ],
    "dataSize": 2
}