Full Name
Tony Jones
Job Title
Principal Technologist
Company
MediaKind
Speaker Bio
Tony Jones has advanced the cutting edge of digital video technology for the past 30 years. Starting in R&D designing post-production digital video effects technology at Questech, he moved to digital transmission systems for satellite, cable, IPTV and OTT, through Tandberg Television, Ericsson and now MediaKind. As Principal Technologist, he currently reports to the Chief Business Officer for MediaKind and plays a key role defining core technology for the portfolio from video processing to TV platforms, including high dynamic range, cloud technology in the media space and end-to-end storage and ABR delivery optimization. Jones received a BSc(Eng) degree from Imperial College of Science and Technology in London and has had nine patents granted relating to advanced digital media.
Speaking At
TTC 2021 Speaker Type
Presenter
Topic and Description
Presentation Topic: Machine Learning for Up-Conversion
Description: Ultra High Definition (UHD) services are progressively being rolled out around the world, but there remains a lack of native UHD content, either because the content was produced before UHD’s standardization, or because the costs were deemed too high. This means that there is a need for conversion from HD to UHD, to fill the gap in content. There are traditional ways to approach up-conversion, but in this presentation we will look at how Machine Learning can be applied successfully to up-conversion, with the aim of producing content that has more native UHD characteristics than traditional approaches might give. The presentation will cover specific architectures of convolutional neural networks (CNNs), with a particular training architecture, known as Super-Resolution Generative Adversarial Networks (SRGANs), and why this provides an enhanced up-conversion. The presentation will also consider interlace, and SD to HD conversion, with some examples of the conversion and lessons learnt from working with various types of content.
and
Roundtable Topic: Machine Learning for Up-Conversion
Description: Q & A Session
Description: Ultra High Definition (UHD) services are progressively being rolled out around the world, but there remains a lack of native UHD content, either because the content was produced before UHD’s standardization, or because the costs were deemed too high. This means that there is a need for conversion from HD to UHD, to fill the gap in content. There are traditional ways to approach up-conversion, but in this presentation we will look at how Machine Learning can be applied successfully to up-conversion, with the aim of producing content that has more native UHD characteristics than traditional approaches might give. The presentation will cover specific architectures of convolutional neural networks (CNNs), with a particular training architecture, known as Super-Resolution Generative Adversarial Networks (SRGANs), and why this provides an enhanced up-conversion. The presentation will also consider interlace, and SD to HD conversion, with some examples of the conversion and lessons learnt from working with various types of content.
and
Roundtable Topic: Machine Learning for Up-Conversion
Description: Q & A Session
