Azure Cognitive Services now ready to run from Docker containers

Microsoft announced that Azure Cognitive Services is now supported by Docker containers, in a bid to make it easier for developers to create machine learning programs. In a batch of releases, the Azure team shared details about the container updates, technical considerations and current use cases by companies.

Container updates

Three Text Analytics containers were highlighted in Microsoft's announcement. Key Phrase Extraction can spot key phrases in inputted text, for instance combing a restaurant or product review. Language Detection is tooled to identify 120 languages and outputs a code for each document that is uploaded, guiding translation efforts. The Sentiment Analysis API ranks sentiments in uploaded documents on a zero to one scale, where one is the most positive.

According to Lance Olson, Microsoft's director of program management for applied AI, containerized instances function very similarly to Cognitive Services APIs in hosted Azure endpoints.

The Face Container is geared toward emotion analysis, detecting and verifying faces. It comes with a common configuration framework allowing customers to tweak telemetry, security, logging, storage and other configuration needs. The Recognize Text Container can extract text data from business cards, posters, receipts and other printed items. Similarly, Custom Vision Service is adding support for logo detection as a way of analyzing social media feeds, even for obscured and partial images.

Customers can use the same APIs to call Cognitive Services or containers. The pricing model is set so that 1,000 API calls for Key Phrase Extraction will cost the same regardless of whether the call is to the API or Cognitive Services. To calculate costs, the containers must be able ...