Pressmeddelande -
Open-source toolkit to automate a data-driven process defining an ABR ladder
Sustainability in streaming will be even more important going forward, and to facilitate bandwidth saving measures, Eyevinn is open sourcing their toolkit they are using when helping their customers to optimize their ABR ladders.
In the early days of streaming services, the highest bitrate was used as a measure of video quality. The higher bitrate, the better quality, was the perception by many consumers. Today we in the industry know that this is not a sustainable approach. Higher bitrate takes more bandwidth of the Internet network resources and to just continue to add more network resources is not the right way to go.
Instead, today the aim is to improve the video quality for the consumers without any unnecessary increate of bitrate. One approach is based on a method called VMAF (Video Multimethod Assessment Fusion). A method that seeks to reflect the viewer’s perception of quality. VMAF gives you a score for how close to the original the down-transcoded copy is. A score above 93 means that a human cannot visually see any difference. The VMAF method is built on research by Netflix and Prof. C.-C. J. Kuo and his group at the University of Southern California.
The process to define an optimal ABR ladder can be a bit tedious and manual [1], as it means we first need to create a variety of videos at different bitrates at each resolution we wish to support. Then we need to measure the video quality for each of these videos to be able to plot the graph that is then used as the decision-basis for the ABR-ladder definition.
To automate parts of this process Eyevinn has developed a toolkit called AutoVMAF [2] that is now available as open source for anyone to download and use. Transcoding process and measurements can be running on on-prem hardware or on AWS using MediaConvert and Elastic Container Service. The toolkit also provides a REST API to orchestrate the transcoding and measurements.
Eyevinn’s VP R&D, Jonas Birmé, gives the following answer to the question why they are open-sourcing it:
“I think there are many streaming services out there that still might be stuck with ABR-ladders defined a couple of years ago based on a general recommendation. By making our AutoVMAF toolkit available as open source we hope to facilitate service providers to use a data-driven process to review and improve the ABR-ladder currently in use.”
Instructions on how to download, install and getting started are available on the online documentation site for AutoVMAF [2].
About Eyevinn Technology
Eyevinn Technology is world leading independent experts in sustainable video streaming technology. For more information about this press release contact:
Jonas Birmé, VP R&D Eyevinn Technology
E: jonas.birme@eyevinn.se
Links:
[1] https://dev.to/video/automating-video-analysis-to-cut-your-streaming-bandwidth-usage-in-half-5hk1
[2] https://autovmaf.docs.eyevinn.technology
Relaterade länkar
Ämnen
Kategorier
Regioner
Eyevinn Technology är det världsledande oberoende konsultbolaget specialiserade på videoteknik och streaming. Med vår expertis hjälper vi vår bransch att säkerställa att användarens upplevelse är av högsta möjliga kvalitet. Vårt oberoende garanterar att våra kunder får opartiska råd och lösningar bäst anpassade för deras unika förutsättningar. Vi erbjuder tjänster inom teknikstrategi, systemarkitektur, projektledning och mjukvaruutveckling.
Eyevinn Technology ingår i Vinngroup, en av Sveriges mest snabbväxande konsultgrupper. Verksamheten går ut på att bygga och utveckla specialiserade företag inom tydliga nischer. Vinngroup får människor och företag att växa och målet är att skapa tydliga resultat hos våra kunder.
