NEW REFERENCE ARCHITECTURE: Batch scoring on Azure for deep learning models

This post has been republished via RSS; it originally appeared at: Azure Global articles.

First published on MSDN on Nov 06, 2018
Written by JS Tan from AzureCAT. Reviewed by Mike Wasson.

• Batch scoring on Azure for deep learning models

Reference architectures provide a consistent approach and best practices for a given solution. Each architecture includes recommended practices, along with considerations for scalability, availability, manageability, and security. This architecture includes a deployable solution as well. The full array of reference architectures is available on the Azure Architecture Center .



This reference architecture shows how to apply neural style transfer to a video, using Azure Batch AI. Style transfer is a deep learning technique that composes an existing image in the style of another image. This architecture can be generalized for any scenario that uses batch scoring with deep learning.

This architecture consists of the following components:

• Compute: Azure Batch AI
• Storage: Blob storage
• Trigger/scheduling: Azure Logic Apps , Azure Container Instances , DockerHub
• Data preparation: AzCopy , FFmpeg

This article covers the following topics:

1. Performance considerations
   1. GPU vs. CPU
   2. Parallelizing across VMs vs cores
   3. Images batch size per Batch AI job
   4. File servers
2. Security considerations
   1. Restricting access to Azure blob storage
   2. Data encryption and data movement
   3. Securing data in a virtual network
   4. Protecting against malicious activity
3. Monitoring and logging
   1. Monitoring Batch AI jobs
   2. Logging in Batch AI
4. Cost considerations
5. Deploy the solution

Head to the article page to learn more and to deploy the solution.



 
AzureCAT Guidance  
"Hands-on solutions, with our heads in the Cloud!"