AWS re:Invent 2017: Serverless Encoding at Scale with Content Moderation via Deep Le (ARC311)

Опубликовано 29 ноября 2017, 16:48

With more companies entering the OTT market, AWS sees customer demand for ways to decrease the time it takes to get content into their users’ hands, while increasing operational efficiency and lowering IT infrastructure costs. Using deep learning-based image analysis can provide users actionable feedback about the content they view. When combining a new serverless architecture approach using Amazon Elastic Transcoder with AWS’ deep learning technology Amazon Rekognition, companies can provide near real-time, on-demand encoding of assets and content moderation. This session covers serverless versus virtualized infrastructure, handling encoding jobs with AWS Lambda, encoding dynamic media assets with Elastic Transcoder (or Elemental), moderating content with Amazon Rekognition, and storing metadata with Amazon DynamoDB. We also provide a demo to test a production-ready serverless encoding architecture.