NASCAR is planning to migrate 18 petabytes of archive video footage to Amazon Web Services, utilize machine learning and ultimately provide fans with over-the-top content and services.
Steve Stum, vice president of operations and tech production at NASCAR, said the company has footage dating back to the 1920s. "We started to digitize six or seven years ago," said Stum. "We are jumping into machine learning to automate processes and add metadata."
Nevertheless, 18 petabytes of data is a heavy lift that equates to 500,000 hours of content.
NASCAR chose AWS as its standard for cloud machine learning and artificial intelligence workloads. NASCAR will use AWS to automate processes related to new video series as well as historical footage. AWS, Google Cloud and Microsoft Azure are all competing for media and video heavy workloads to highlight their AI tools.
Specifically, NASCAR will use Amazon Rekognition to tag video with metadata such as driver, race, lap time and sponsors. The metadata will be used to deliver video footage on NASCAR.com and social media. NASCAR will also use Amazon SageMaker to train deep learning models against 70 years of footage.
Stum said the goal is to be able to use machine learning to spot NASCAR executives and drivers. "The goal is to see Richard Petty in the 1960s and recognize him today," he said.
According to Stum, NASCAR isn't expecting a lot of savings from automating tagging and added that workers will be retrained to oversee machine learning processes.
Stum added that NASCAR is working on 10GbE connections to move its 18PB archive to AWS. New video footage will move to AWS directly, but archives will take until the end of the year to move completely.
Once all the video is in one place, NASCAR will combine databases for machine learning combine metadata. NASCAR has done high level testing and found machine learning is faster and more accurate when it comes to tagging content.
Stum said that NASCAR looked at multiple cloud providers but was comfortable with AWS since it was already using the Amazon Media Services, AWS Elemental MediaLive and MediaStore.
Correction: A previous version had Steve Strum. The correct spelling is Steve Stum.