Netflix’s Anne Aaron explains how VMAF came about and how AV1 is going to benefit both the business and the viewers.
VMAF is a method for computers to calculate the quality of a video in a way which would match a human’s opinion. Standing for Video Multi-Method Assessment Fusion, Anne explains that it’s a combination (fusion) of more than one metric each harnessing different aspects. She presents data showing the increased correlation between VMAF and real-life tests.
Anne’s job is to maximise enjoyment of content through efficient use of bandwidth. She explains there are many places with wireless data is limited so getting the maximum amount of video through that bandwidth cap is an essential part of Netflix’s business health.
This ties in with why Netflix is part of the Alliance for Open Media who are in the process of specifying AV1, the new video codec which promises bitrate improvements over-and-above HEVC. Anne expands on this and presents the aim to deliver 32 hours of video using AV1 for 4Gb subscribers.
Netflix has famously moved in to original content but less-known are its innovations behind the scenes in production workflows.
Eric Reinecke looks at the challenges in moving media and finding ways to correctly pick and choose the right media to move. He looks at the different ways of moving editorial data: the venerable EDL, Avid’s more recent AAF and Final Cut’s XML talking about the pros and cons of them all.
The talk then moves on to OpenTimelineIO which is an API and interchange format for editorial cut information which was designed to help departments in animation studios to work together. Hosted by Pixar, companies like Netflix are finding uses for the API outside of animation and Eric shows demos of how he’s using it within Netflix then ends with a call to get involved!
To mark International Women’s Day these videos are free to watch, but for 3 days only! Free registration required.
To mark International Women’s Day, Eyevinn Technology have opened their premium archives to allow you to watch two videos for free until 11th March which are normally reserved for patrons of the Streaming Tech Sweden conference – so act quickly to watch Netflix’s Megha Manohara discuss Netflix’s dynamic optimiser framework and how they ensure best quality over a variety of bandwidths.
Megha covers encoding testing, metrics such as VMAF, visualising the results, per-shot encoding and the way they validate with their audience they have done a good job.
Streaming Tech Sweden is an annual conference which prides itself on excellence and independence. Without sponsors, they are free to pick the best and the most relevant speakers working on at the cutting edge of video streaming. The talks from Streaming Tech Sweden 17 are free to watch, but those from 2018 are available for attendees only. Later in 2019, they will become free, but until then, this is a short opportunity to watch these two great talks in order to mark International Women’s Day 2019. Registration with the site is free.
The second talk available from Streaming Tech Sweden 2018 is from Codemill’s Johana Björklund talking about contextual marketing and ad personalisation. Johana explains how the ads work, how GDPR has changed the way personalisation is carried out and how video metadata is used to find pre-roll and post-roll ads.
VMAF is a video quality metric created by Netflix which allows computers to indicate what quality a video is. This is an important part of evaluating how good your encoder or streaming service is so it’s no surprise that Netflix has invested years of research into this. Other metrics such as PSNR and MS-SSIM all have their problems – and let’s accept that no metric is perfect – but what the industry has long grappled with is that a video that has a strong fidelity to the source doesn’t necessarily look better than one that less-faithfully replicates the source.
Imagine you had a video of an overcast day and one encoder rendered the video a bit brighter and a bit more blue. Well, for that clip, people watching might prefer that encoder even though the video is quite different from the source. The same is true of noisy pictures where replicating the noise isn’t always the best idea as some people, for some content, would prefer the cleaner look even though some details may have been lost.
As such, metrics have evolved from PSNR which is much more about fidelity to metrics which try harder to model what ‘looks good’ and VMAF is an example of that.
Zhi Li explains the history of VMAF and explains some of the new features which were released in August 2018, when this talk was given, which gives an insight into the way VMAF works. Plus, there’s a look ahead at new features on the road map. This talk was given at a SF Video Technology meet up.