Container support for Cognitive Services APIs

  • Article

[日本語版はこちら]

Now Azure Cognitive Services Containers is available for preview, which enables you to deploy Cognitive Services in Docker container (including local container). It means you can use cognitive features in your own containers, be free from delay due to internet connection, or get feature to keep your data stored locally.

Microsoft Azure Blog > Bringing AI to the edge Microsoft Azure Blog > Getting started with Azure Cognitive Services in containers Microsoft Azure Updates > Azure Cognitive Services Containers are in preview

Compatible Docker Environments

You can use these Docker environments:

  • Local on Windows*/macOS/Linux (*be configured to support Linux containers on Windows)
  • Azure Kubernetes Service
  • Azure Container Instances
  • Kubernetes cluster deployed to Azure Stack

Available Cognitive Services features in containers

Available Cognitive Services features in containers are:

  • Computer Vision (sign-up for preview request)
    • OCR - Recognize Text Extracts printed text (currently works English only)
  • Face (sign-up for preview request)
    • Face Detection
      • with attribute identification such as face landmarks (noses, eyes, ...), gender, age, etc
    • Face Verification, Face Identification
  • Text Analytics (now available)
    • Key Phrase Extraction
    • Language Detection (120 languages)
    • Text Analytics Sentiment Analysis

Process to use Cognitive Services Containers

To use Cognitive Services Containers,

  1. Sign-up for preview request (Computer Vision Recognize Text & Face)
  2. Create Cognitive Services on Azure Portal
  3. Download Cognitive Services container image, instantiate a container

details: Cognitive Services Face API Documentation > Install and run containers

now you can access your Cognitive Services (ex: Face AI, hosted port 5000) using endpoint such as

 POST https://localhost:5000/face/v1.0/detect

intead of accessing via (current) Web API endpoint:

 POST https://westus.api.cognitive.microsoft.com/face/v1.0/detect

Access as Rest API, or you can still use same Cognitive Services Client Libraries. Enjoy!