OCR for Brazilian CNH (driver's license) to JSON object. It will try to read all the driver's license data, and allows you to compare/match by the similarity of a selfie against the document photo.

The simplest way to get started with this project in 2 steps:

docker build -t ocr-cnh .

It may take a while as the Dockerfile compiles and install several OS native libraries to run the service.

After the building, just run an instance of the container by executing:

docker run -it -d -p 8080:8080 ocr-cnh

You can check if the container is running accordingly by executing the docker ps. Don't forget take a look at the logs (docker logs {container_name_or_id})

At this point you're ready to go. Just access http://localhost:8080 and you should see a nice and friendly form to test it :)

As said, this is a very simple service, with a even simpler use dynamic:

• You send (upload within a POST multipart/form-data request) the document (CNH) image;
• The service will return you a JSON with the document information (or not - it depends on the quality of the document and image)

The following code snippet examplifies how to use the API:

import fetch from 'fetch';

let formData = new FormData();
let fileField = document.querySelector('input[type="file"]');
let file = fileField.files[0]

if (!file) alert('ERROR: Select a file.')

formData.append('file', file);

try {
    const rawResponse = await fetch('/cnh.json', { method: 'POST', body: formData });
    const response = rawResponse.json();
    console.log(response);
} catch (error) {
    console.error(error);
}

This request will return a JSON with the following format:

{
	"nome": "XXXXXXXXXXXXXXX",
	"cpf": "XXXXXXXX",
	"dt_nasc": "XXXXXXXX",
	"rg": "XXXXXXXX",
	"rg_emissor": "XXXXXXXX",
	"rg_uf": "XXXXXXXX",
	"numero": "XXXXXXXX",
	"cidade": "XXXXXXXX",
	"uf": "XXXXXXXX",
	"pai": "XXXXXXXX",
	"mae": "XXXXXXXX",
	"emissao": "XXXXXXXX",
	"validade": "XXXXXXXX",
	"avatar": "XXXXXXXX"
}

Contributors.