Quantum NVIDIA AI and ML Content Enhancement


Category  Artificial intelligence AI - Machine learning ML

DMW Awards Finalist

Computer machine learning is a powerful technology with the potential to tirelessly enhance and enrich content as it flows through a content production workflow, and across deep content repositories common for continuous content producers for broadcast, sports, television, and film production. However, integrating AI and ML models into production and content monetization workflows can be highly complex – and even impractical if content needs to be sent to the cloud for every operation. The ideal solution should leverage on-site hardware for maximum performance and control and be tightly integrated with familiar content collaboration frameworks, such as media asset management solutions for the highest efficiency. Above all, any new AI and ML solution should not be disconnected islands of work. They should amplify and automate the work of your content experts to speed production time and make content discovery and enhancement across even the largest content libraries possible, freeing logging staff and creatives to focus on higher level creative decision making.

The Solution

With this in mind, Quantum has developed the Quantum AI and ML Content Enhancement Solution. Powered by the Quantum CatDV asset management and workflow orchestration platform and the Quantum StorNext® shared storage and file archive and tiering engines, Quantum has integrated powerful NVIDIA AI/ML SDK’s and high-performance NVIDIA GPU systems to deliver a tightly integrated asset management and content enhancement solution.

Within the familiar CatDV interface, customers can now initiate or even automate powerful ML-powered actions directly within their content collaboration workflows to dramatically speed laborious tasks, such as logging. Customers can automate the creation of clips of interest for human review, such as ‘compose an asset of sub-clips from a match that shows a goal with a 15s lead-in and lead-out.' All markers created by the solution are fully searchable, bringing users directly to points of interest within assets for further review.

Common tasks, such as object recognition within frames of video, speech-to-text translation, video and audio super-resolution, and other tools add valuable metadata to video and image files, saving time and money and unlocking hidden value in media content. Many actions are performed faster-than-real-time thanks to the direct access to NVIDIA GPU processors for rendering, transcoding, and machine learning tasks. Beyond the initial set of machine learning tasks, customers are offered guidance and assistance in training custom solutions that match their content subject matter and specific needs.

The solution includes:

- CatDV Enterprise solution with CatDV Automation Workers.

- AI/ML integration based on NVIDIA DeepStream, Riva, and Maxine SDKs.

- NVIDIA GPU hardware pipeline integration for CatDV proxy generation, transcoding, and AI acceleration.

- Additional optional custom AI/ML development to customer needs or to enable customer’s development teams on the common platform.

- StorNext environment integration to customer needs.

- Optional NVMe storage, such as F2100 to accelerate content movement within the GPU pipeline.

- Optional archive to ActiveScale, Scalar tape, or S3-compatible storage per customer needs.