French ID extraction

This workflow parses images of French identity documents and returns structured identity fields.

Capabilities

• Extracts informations from supported French identity documents : passports, national identity cards, driving licences, and residence permits.
• Supports images containing more than one document and returns one result object per detected document.
• Automatically handle document detection and corrects orientation.

Supported documents

• French passport: PASSEPORT.
• French national identity card, old format: RECTO_CNI_OLD, VERSO_CNI_OLD.
• French national identity card, new format: RECTO_CNI_NEW, VERSO_CNI_NEW.
• French driving licence, old format (folio): RECTO_PERMIS_OLD, VERSO_PERMIS_OLD.
• French driving licence, new format (card): RECTO_PERMIS_NEW, VERSO_PERMIS_NEW.
• French residence permit: RECTO_TS, VERSO_TS.

Output fields

The output is a list of detected documents. Each item can contain:

• document_type
• document_detection_confidence
• information_extraction_confidence (absent for detection-only results such as VERSO_CNI_OLD)
• raw_information: fields predicted by the extraction model before post-processing.
• computed_information: post-processed fields with various cleaning and normalization.
• image: base64-encoded cropped document image (when parameter encode_img is True).

Fields in raw_information and computed_information:

• last_name
• given_name_1
• given_name_2
• given_name_3
• given_name_4
• alternate_name: alternate, usage, or spouse name when present
• birth_date
• birth_place
• gender
• nationality
• country
• document_number
• card_access_number
• issue_date
• issue_place
• expire_date
• license_categories: driving licence categories
• obtention_date_b0: obtention date for the B0 driving licence category, when present
• expire_date_b0: expiry date for the B0 driving licence category, when present
• mrz1: first MRZ line
• mrz2: second MRZ line
• mrz3: third MRZ line

Not every field is expected for every supported document. Fields that are not present on a given document may be returned as empty strings.

Known limitations

• Only French identity documents are supported
• Old French identity card backs (VERSO_CNI_OLD) are detected but no extractions are performed on them.
• Extraction quality depends on image resolution, sharpness, lighting, glare, document visibility

Usage

Python

1. First, deploy the workflow.
2. Install the Python library:

pip install ikomia-client

3. Integrate the workflow into your Python code:

from ikclient.core.client import Client

  with Client(
      url="https://your.scale.endpoint.url",
      token="your-api-token",  # Defaults to os.environ.get("IKOMIA_TOKEN")
  ) as document_extractor:
      results = document_extractor.run(
          "path/to/your/document.jpg",
          parameters={
              "encode_img": False,
          },
      )

      output = results.get_output()
      extracted_documents = output.data

      for document in extracted_documents:
          print("Document type:", document.get("document_type"))
          print("Detection confidence:", document.get("document_detection_confidence"))
          print("Extraction confidence:", document.get("information_extraction_confidence"))
          print("Computed information:", document.get("computed_information"))

For more details, refer to the documentation.

JavaScript

1. First, deploy the workflow.
2. Install the JavaScript library:

npm install @ikomia/ikclient

3. Integrate the workflow into your JavaScript code:

import { Client, ImageIO } from "@ikomia/ikclient";
  import fs from "fs/promises";

  const client = new Client({
    url: "https://your.scale.endpoint.url",
    token: "your-api-token", // Defaults to process.env.IKOMIA_TOKEN
  });

  const imageBuffer = await fs.readFile("path/to/your/document.jpg");
  const imageInput = await ImageIO.create(imageBuffer);

  const results = await client.run({
    inputs: [imageInput],
    parameters: {
      encode_img: false,
    },
  });

  const output = results.getOutput();
  const extractedDocuments = output.data;

  for (const document of extractedDocuments) {
    console.log("Document type:", document.document_type);
    console.log("Detection confidence:", document.document_detection_confidence);
    console.log("Extraction confidence:", document.information_extraction_confidence);
    console.log("Computed information:", document.computed_information);
  }

For more details, refer to the documentation.

REST API

Refer to the documentation for integration using the REST API.