Nothing quite beats the shared experience of trade events and conferences. The value of experts and enthusiasts coming together from across the industry to knowledge share and network – from students and academics to industry veterans and specialists – cannot be overstated. It was a pleasure attending ACM Mile High Video 2022 (MHV22), which took place in Denver, Colorado, earlier this month. The video coding and streaming conference was one of the first large-scale events focusing on media distribution since the pandemic. It was fantastic to see such a strong turnout – 275 attendees across the video coding industry were there, with a further 1025 viewers tuning in remotely to watch the event’s live stream.
With three keynote speakers and three tutorials from Emmy-award-winning industry veterans, the event offered some fascinating updates and overviews on some of the latest industry hot topics and video coding standardization processes. One interesting presentation that stood out for me was from Benjamin Bross (Fraunhofer Heinrich Hertz Institute); Mathias Wien, Jens-Rainer Ohm (RWTH Aachen); Gary J. Sullivan (Microsoft Research); and Yan Ye (Alibaba), where they provided an update on the emerging Versatile Video Codec (VVC) standard and its applications. Amaya Jiménez-Moreno, Lorenzo Ciccarelli, Rick Clucas and Simone Ferrara from V-Nova also provided an interesting overview of LCEV, which was recently standardized by MPEG and ISO as MPEG-5 Part 2. LCEVC has just been adopted by the next generation Brazilian digital television standard, an important step towards a widespread adoption of this newly proposed standard.
In “Video Codec Licensing Landscape”, Carter Eltzroth from Helikon focused on how codec licensing is changing in the wake of the controversial HEVC licensing model, which was not only detrimental to a wider adoption of the standard but also motivated the emergence of concurrent codecs.
In “Video Streaming Engineering: Our Industry Learns from Failure”, David Hassoun and Jun Heider from Dolby Labs reviewed some of the hard lessons they learned while deploying large-scale streaming systems. This clearly illustrates the collaborative spirit of the event, where knowledge sharing is seen as the best way to make the industry thrive and create new opportunities for all players.
AI and ML – A Hot Topic!
More companies are realizing the potential of Artificial Intelligence (AI) and Machine Learning (ML) within media delivery, which was evident at MHV22. Being an area where MediaKind has spent a lot of time researching and investing and was awarded a Technical Emmy for our work in “AI/Optimization for Video Compression: Improving perceptual quality metrics and their application to real-time and non-real-time video compression” last year, it was brilliant to share our experiences and innovations in the space at MHV22 while hearing updates from a broad spectrum of voices across the media and content industry.
COMCAST and Media Distillery shared how they are using AI to generate highlights or for topic detection, enhancing the content discovery experience for the viewers and increasing possibilities for target advertisement. Start-up company Small-Pixels also shared how they’re proposing Super-Resolution upscaling and compression artifact removal using AI at the decoder side to reduce bandwidth by transmitting video at a lower resolution. Researchers from the Alpen-Adria Universität Klagenfurt also proposed a similar approach for low-power mobile device applications. Another start-up, wave.one, shared how it is proposing full native video encoding using neural networks to exploit a whole new paradigm in video compression.
I was glad to share a presentation on the AI and ML topic myself, exploring MediaKind’s solution for Optimizing Real-Time Video Encoders with ML. The presentation focused on how MediaKind has developed innovative approaches to adapt both high-level and low-level encoding strategies on real-time encoders with limited computing resources. By estimating the likelihood of different mode decisions and the impact of different encoding tools in the encoding process, we can more efficiently allocate the computational resources to be used in the way that most benefits compression efficiency and video quality.
The advantages are clear – maximizing the video quality or further reducing computational requirements! We adopted a pragmatic approach where low complexity was prioritized over accuracy so that our solution could be used in a wide range of applications. By using Deep neural networks that can be implemented efficiently in CPUs, our solution doesn’t require any specific hardware and can be deployed in any cloud provider.
You can learn more about the significant savings this brings on hardware and energy costs, with clear environmental benefits, here.
Uniting the opensource community
MHV22 was also used to showcase progress made in the opensource community. Fraunhofer FOKUS shared its latest advances in developing the Open-Source Player dash.js. Netflix shared its progress in the newly proposed Contrast Aware Multiscale Banding Index, a metric designed to provide a banding artifact evaluation and that has been made available as part of their VMAF video quality metric library. VMAF was jointly developed by Netflix and several academic collaborators, becoming a widely used and accepted opensource standard for evaluating the performance of encoding systems and driving encoding optimizations.
MHV22 was a huge success, and I’m incredibly thankful to Dan Grois, Christian Timmerer, Alex Giladi, the show’s sponsors, and everyone else involved in organizing and promoting this event and bringing the video coding and streaming industry together. It was an invaluable opportunity for start-ups to showcase their innovation to larger, more established companies. The mutual benefit was clear – start-ups could reach new potential customers and grow their business, while established players tapped into the latest products and technologies within the industry, unlocking new markets and applications to increase their portfolio offerings.
All of the presentations will be made available on MHV’s YouTube page.