This is the second post in an exclusive Avid series on artificial intelligence (AI) and new technologies. Explore Avid’s RADLab (Research and Development) and how we are tapping into the power of emerging technologies to innovate and support our customers. Visit our AI and emerging tech resource page for more on this hot topic.
At Avid, we have always immersed ourselves in emerging technologies so that we can provide our customers with the most current, most efficient solutions possible. To help us understand how best to utilize new technologies like artificial intelligence (AI), we established an in-house research lab dedicated to bringing innovation forward at an accelerated pace.
To shed some light on the Avid RADLab the groundbreaking work it’s doing, and how that translates to supporting our customers and the media industry, we talked with two Avid media tech experts: Shailendra Mathur, Vice President of Architecture and Technology, and Rob Gonsalves, Engineering Fellow and AI enthusiast.
Q. What’s the RADLab and how is it set up to explore the use of AI in media production?
A. The Avid RAD (Research and Development) Lab was set up in 2021 to explore the use of new technologies, including AI, in media production. The lab takes on research projects to feed our product innovation engine. Our RAD team includes Avid’s technology and business leaders, graduate student interns studying computer science and machine learning, and in some cases specialized development partners. Our interns and contractors come from many countries, bringing diverse backgrounds and perspectives that positively shape our approach to research.
In addition to helping students progress in their academic careers, the RADLab has a broader goal of supporting and collaborating with the industry. We actively seek collaboration with others, sharing our work to advance the field overall.
Q. There’s so much happening in AI and new technologies! How do RADLab leaders decide what research to pursue?
A. The RADLab follows a multistage process where topics are selected and prioritized based on practicality, innovation, and technical feasibility. If the research proves the technology’s viability and aligns with business needs, the RADLab then develops a prototype. Failure is absolutely an option. We approach this as a fail-fast-succeed-fast effort. Even failures are useful learning – sometimes to then try a new approach, or other times to discover why hype does not meet reality or practicality.
The RADLab research would be useless if we did not get it in the hands of customers. Happily, the more practical and feasible projects that have high business value are now undergoing productization in some of our products. RADLab is about looking at how we make new technologies practical for our customers. It’s a proving ground.
Q. Several interns had a significant role in the RADLab and even had their research published. Tell us about that.
A. Avid is facing the same challenge that the rest of media industry is – competition for young talent and attracting them to the media space. The RADLab has served to attract that talent to our industry, allowing us to hire some of them after they have graduated to help productize what they helped research. The interns are actually very qualified researchers who are focused on their master’s or PhDs. We have been very fortunate to get their fresh thinking and energy (and their professors’ expertise as well) contributing to our problem domains. In return they get exposure to the exciting challenges the media and entertainment market provides.
It is a goal to share our research with the industry. This also gels well with the need of the interns to earn their respective degrees. As examples, two of our interns, Zahra Montajabi and Vahid Khorasani Ghassab, wrote academic papers about their work in the April 2023 SMPTE Motion Imaging Journal. (You can access these articles if you join SMPTE. It’s a great organization of talented people working in media and entertainment worldwide.)
Led by Zahra, the RadLab team sought to solve a complex issue in the field of image processing— identifying the most significant areas of an image. They developed a method to detect objects in an image and perform a detailed search for these objects based on their meaning and relevance. The identification of these critical areas can greatly improve the visual quality of images and videos. The team’s article, “Semantic Regions of Interest Applied to Adaptive Image Compression, Color Enhancement, and Automated Pan-and-Scan,” appeared in the SMPTE journal.
Vahid and team developed a new method for compressing video files, inspired by the fact that human eyes are more sensitive to differences in light than in color. Instead of treating luma and chrominance information equally, as is typically done, they used more luma than chrominance information. They also avoided upscaling video to a larger format, reducing computational load. Their method improved video quality by five to six percent and decreased the computational effort by 37 to 40 percent. Their approach could enhance current video compression techniques. The team’s article, “Optimizing Video Compression with CNN-Based Autoencoders with Chroma Subsampling,” also appeared in the journal.
Find out more about Avid’s involvement in AI and the anticipated impact of new technology on media production.
