Guide for Integrating Private AI DEID container with Azure OCR

This guide provides step-by-step instructions on setting up the Private AI container with the Azure OCR for processing images and documents. Both the cloud and containerized versions of Azure OCR are supported.

Prerequisites

Docker Compose installed on your system.
Access to Private AI's DEID container. See Grabbing the Image for more info on this.
An Azure account and a subscription to Azure Computer Vision service.

Step 1: Create Azure Computer Vision Service

Navigate to the Azure Portal.
Create a new Azure Computer Vision service by following the link: https://portal.azure.com/#create/Microsoft.CognitiveServicesComputerVision . (You can also use your existing Azure Computer Vision service)
Note down your Computer Vision service endpoint and the key.

Step 2: Configure Docker Compose File

Create a docker-compose.yml file in your preferred text editor.
Configure the DEID service and Azure OCR either using either cloud (2A) or containerized (2B) Azure Computer Vision service.

Step 2A: Docker Compose for Cloud Azure Computer Vision Service

Copy

Copied

version: '3.8'
services:
  deid_service:
    image: crprivateaiprod.azurecr.io/deid:gpu
    shm_size: "4g"
    container_name: deid_container
    tty: true
    environment:
      - PAI_OCR_SYSTEM=azure
      - PAI_AZ_COMPUTER_VISION_URL=<Azure Computer Vision Service URL>
      - PAI_AZ_COMPUTER_VISION_KEY=<Azure Computer Vision Key>
      - PAI_OUTPUT_FILE_DIR=<your output directory>
    ports:
      - "8080:8080"
    # Input and output directories are only required when using URI file processing
    volumes:
      - <path to license.json>:/app/license/license.json
      - <input file directory>:<input file directory>
      - <output file directory>:<output file directory>
    # comment out this section if you're using a CPU version of deid container.
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

Replace placeholders (e.g., <Azure Computer Vision Service URL>) with your actual service details.

Step 2B: Docker Compose for Containerized Azure Computer Vision Service

Copy

Copied

version: '3.8'
services:
  deid_service:
    image: crprivateaiprod.azurecr.io/deid:gpu
    shm_size: "4g"
    container_name: deid_container
    tty: true
    environment:
      - PAI_OCR_SYSTEM=azure
      - PAI_AZ_COMPUTER_VISION_URL=http://azure_vision_service:5000
      - PAI_AZ_COMPUTER_VISION_KEY=<Azure Computer Vision Key>
      - PAI_OUTPUT_FILE_DIR=<your output directory>
    ports:
      - "8080:8080"
    # Input and output directories are only required when using URI file processing
    volumes:
      - <path to license.json>:/app/license/license.json
      - <input file directory>:<input file directory>
      - <output file directory>:<output file directory>
    # comment out this section if you're using a CPU version of deid container.
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

  azure_vision_service:
    image: mcr.microsoft.com/azure-cognitive-services/vision/read:latest
    container_name: azure_vision_container
    tty: true
    environment:
      - Eula=accept
      - Billing=<Azure Computer Vision Resource Endpoint>
      - ApiKey=<Azure Computer Vision Resource Key>
    deploy:
      resources:
        limits:
          memory: 16G
          cpus: '8'

For further details on the Azure Computer Vision container please refer to the Azure Documentation.

Step 3: Start the Containers

Open a terminal and navigate to the directory containing your docker-compose.yml file. Run the following command to start the containers:

Copy

Copied

docker compose -f docker-compose.yml up

Step 4: Verify Operation

Once the containers are running, you can verify their operation by accessing the DEID service's exposed port (e.g., http://localhost:8080) and performing a test OCR operation on your documents or images.