Project description
AI mimics human eye for ultra-efficient video compression
The explosion of data and our desire to connect everything, from our ovens to self-driving cars and surgical equipment, to the internet is causing a tremendous flurry of research into how to accommodate all this fast data transmission reliably, securely and with low power consumption. A significant portion of research is directed at technologies to increase bandwidth. The EU-funded AISTREAM project will offer an innovative solution focused on reducing the data – not the actual data going in and coming out but the size during transmission – via AI-supported compression. The team will focus on video, which accounts for a huge volume of the internet traffic, and will demonstrate the technology in a streaming service.
Objective
Deep Render combines the fields of artificial intelligence, statistics and information theory to unlock the fundamental limits of video compression.
The best data compressor known to mankind is the human eye, with compression ratios at least 2,000 times better than anything developed to date. Our Biological Compression technology mimics the neurological processes of the human eye through a non-linear, learning-based approach, creating an innovative class of highly efficient compression algorithms. By building an entirely new foundation for compression, avoiding the limitations of current codecs, our objective is to develop a video compression approach 80% more efficiency than the state-of-the-art.
With 85% of all internet traffic being video data, growing exponentially, bandwidth supply is being used up at an unsustainable rate. Even worse, emerging video technologies such as VR-streaming, Medical and Satellite Imaging, and Autonomous driving are bottlenecked by the unavailability of sufficient bandwidth. Further, the amount of energy used and CO2 generated with online video is now being recognised as a major problem.
If the EU Digital Single Market and economic growth are to be delivered, and Climate Change obligations met, a more efficient compression system is vital to free up bandwidth and reduce energy usage. Our value proposition is simple, by reducing file sizes by 80%, we directly increase the bandwidth supply of the internet by a factor of 5, thus reducing data transport and storage requirements, reducing energy usage and CO2 emissions.
Initially, the end-users of our technology will be content delivery networks, online streaming services and media production organisations. The video encoding market is estimated to be worth €1.5Bn a year. Our collaboration, including TU Wien and Contentflow (end-user), will develop, demonstrate and pilot the codec in a streaming service and begin extending the codec to new high growth, high value and high need markets.
Fields of science
CORDIS classifies projects with EuroSciVoc, a multilingual taxonomy of fields of science, through a semi-automatic process based on NLP techniques.
CORDIS classifies projects with EuroSciVoc, a multilingual taxonomy of fields of science, through a semi-automatic process based on NLP techniques.
- natural sciencescomputer and information sciencesartificial intelligence
- natural sciencescomputer and information sciencesinternet
- engineering and technologymechanical engineeringvehicle engineeringaerospace engineeringsatellite technology
- natural sciencesearth and related environmental sciencesatmospheric sciencesclimatologyclimatic changes
Keywords
Programme(s)
Funding Scheme
IA - Innovation actionCoordinator
1040 Wien
Austria