November 24, 2021 – A new, improved version of our video restoration algorithm called Pixop Deep Restoration 2 was recently deployed to production. The filter is designed for enhancing digital live action footage and achieves visual results which are on par with the best native software product on the market today. Several months of research and experimentation went into the project.
September 1, 2020 – Five machine learning based video processing algorithms for e.g. denoising, restoration, upscaling via Super Resolution, etc. have been developed, trained and deployed on AWS infrastructure. They are now available as part of Pixop’s services (web app + REST API). All but one of these algorithms are deep CNN based.
October 14, 2019 – An article about video upscaling methods that I authored was published to Medium. The upshot is: Pixop’s video Super Resolution performs the most accurate reconstruction of the algorithms tested.
September 17, 2019 – Pixop was an exhibitor at the IBC show, showcasing our video Super Resolution technology and web app which enables cloud based video enhancement to be done within a web browser. No special hardware or software installation required. Our video demo:
December 31, 2018 – My open source video comparison tool is now up on GitHub. video-compare can be used to visually compare e.g. the effect of codecs and resizing algorithms on two video files played in sync. The tool is not very restrictive as videos are not required to be the same resolution, color format, container format, codec or duration. Pre-built Windows 64-bit releases are available (for other platforms the tool must be built from source).
November 23, 2018 – I am now the full-time CTO of Pixop.
December 23, 2017 – I co-founded the company Pixop with Morten Kolle. We are taking AI/ML video enhancement in the cloud to the next level. Our mission is to help everyone from indie creators to professional post-producers improve the perceived quality of imagery for the viewing pleasure of us all. Looking forward to the journey!
August 18, 2017 – Two videos were uploaded to YouTube today demonstrating what I have been working on the past months: an implementation of Rapid and Accurate Image Super-Resolution (RAISR). This is a very exciting image upscaling algorithm which has many desirable properties – most importantly that the upscaled images are significantly less blurry than those produced by classic interpolation algorithms. High frequencies are clearly better preserved with RAISR upscaling, i.e. edges remain clearly defined and appear sharper.
I will continue development on this technology as there are plenty of issues and tasks still to be resolved. First of all by making an efficient implementation (massive parallelism is a must for video upscaling) and then by creating commercial products. I’m also looking for people to join me on this journey – drop me a message if you are interested and have a passion for image processing and/or machine learning.
March 13, 2017 –Tifoon, a free open source network ports monitoring application with various reporting capabilities has been released on GitHub. Check it out!