Recent Posts



Tapping Into the Power of Machine Learning

by Geoff Stedman, Chief Marketing Officer

Talk to any number of industry pundits or vendors, and you’ll hear them agree that machine learning (ML) has the potential to improve efficiencies and productivity for media organizations. Just how much potential depends on how well ML is used across the media supply chain to analyze media, create time-based metadata, automate processes, and facilitate more focused manual intervention where required.

SDVI understands how ML techniques can be applied to media at various steps in the supply chain to pinpoint specific media locations that require human attention. People can “work smarter” in dealing with manual tasks and dedicate more of their time and effort to higher-value creative work, or simply get more done in less time. The SDVI Rally cloud-native media supply chain platform integrates a variety of ML engines and applies them to tasks that are needed to move content through a supply chain. For example, ML tools such as AWS Rekognition, Google Cloud Platform VideoIntelligence, and Microsoft Azure VideoIndexer can identify scenes with violence, nudity or bad language within a piece of video for content moderation use cases; perform object detection for metadata enrichment; locate all video and audio errors for QC operations; compare incoming metadata to black segments for segment timing; and automatically detect the language of audio tracks for language verification. The time savings and productivity gains for work that is necessary, but not necessarily the highest value or most creative, is one of the key benefits that SDVI delivers to customers.

Rally supports the use of ML unusually well because the platform boasts integrations with a variety of providers, and because it normalizes ML-generated metadata into a common format, making it readily usable — regardless of its source — to optimize tasks across the supply chain. SDVI customers therefore have the freedom to choose the best ML tool (or tools) for any particular job, and the power to aggregate and leverage a data lake of time-based metadata to make searching, QC, compliance, and other library functions faster, more accurate, and more targeted.

Collecting data at every step of a supply chain, including data from ML processes, Rally can summarize and present information that gives operators immediate insight into their supply chain performance. Over time, as Rally accumulates and aggregates supply chain data, the platform will be able to incorporate its own ML models to provide recommendations or suggest optimizations to the operators’ supply chains.

ML isn’t going to replace humans in the supply chain any time soon, but it will make those humans more productive. People won’t need to waste time reviewing hours of content when a machine can do the same in moments, sending up an alert when human review is required. They can stop parsing through different data formats to extract useful metadata and instead take care of tasks only a human operator can complete. Think of the efficiencies that are possible when media moving through a supply chain can be managed on an exception basis only, with automatically generated work orders for only those instances where a ML model has detected a potential problem that needs investigation.

Supply chains that embrace ML to handle mechanical, non-creative workloads can yield both productivity gains and an increase in content throughput. The same team of operators can now process more content or focus their time on more creative work that results in better content.

The resulting operational efficiency is incredibly valuable in today’s marketplace, where a media organization’s ability to process and deliver content quickly can be a competitive differentiator. Delivering better content (and more of it) is an even bigger win. ©2021 by SDVI. SDVI® and SDVI Rally® are registered trademarks. All rights reserved.