How artificial intelligence is transforming the field of video compression
Image and video data comprise most of the information transmitted on the internet and is expected to grow significantly in the coming years. As a result, bandwidth supply is being used up at an unsustainable rate. The amount of energy used and CO2 generated with online video is a major concern, heightening the need for efficient image and video compression methods to sustain communication stability. Deep Render, an R&D start-up specialising in AI-based video and image compression technologies, has developed a novel approach to video compression. It combines AI, machine learning, statistics, and information theory, in a non-linear approach to video compression that mirrors the neurological processing of the human eye. “In the EU-funded AISTREAM project, we aimed to continue to develop the AI-based compression technology. The goal was to prove that the technology was ready for the next step – going to market,” explains Chri Besenbruch, project coordinator and co-founder of Deep Render.
Bringing the AI-based compression to market
“After years of hard work and exceptional research and development, the technology is finally here, promising massive compression performance boosts, and will soon disrupt the global compression markets. It has graduated from an ‘emerging’ technology to a ‘ready for-usage’ technology,” confirms Besenbruch. The ecosystem is also ready. “The AI-accelerator ecosystem has reached a critical point and is now allowing the execution of video AI on mobile/edge devices,” adds Besenbruch. Furthermore, extending AI-based compression to exotic data types (VR video, medical data, satellite data) is straightforward. For the first time, there is an opportunity to create market/data-specific compression technology. The project also confirms that customers are excited about AI-based compression. “They want to try and roll it out as soon as possible. However, to get the technology into production or beta testing, a few streaming-related features that still need to be built are missing,” outlines Besenbruch.
Contributing towards the EU’s climate change mitigation
In the long run, project work – better compression technology and its application to emerging technologies – impacts three key areas. Firstly, it unlocks innovation by improving existing products and enabling new ones, a goal of the EU. It also impacts on sustainability by prolonging the life of existing infrastructure. “Better compression makes file sizes smaller, allowing the existing infrastructure to handle more data. De facto, better compression acts as a multiplier to the existing fibre optic infrastructure. Hence, five times better compression generates around 15 trillion dollars of value for society and minimises infrastructure related distributions to the environment,” reports Besenbruch. What’s more, people are often unaware of the carbon footprint of online video consumption. The emissions of online videos correlate with file size. Larger file sizes mean more data storage and warehousing (cooling, electricity), more data transmission (electricity), and more infrastructure needs (distribution to the environment). “Better compression technology decreases file sizes, thus decreasing emissions linearly. Five times better compression means a reduction in emissions of five times,” highlights Besenbruch. This contributes towards the EU’s climate change mitigation through reduced CO2 generation and helps transition the video on demand sector to a clean economy by significantly reducing its energy and carbon footprint. Lastly, there is a strong correlation between economic growth and the growth in bandwidth supply, as newer digital services typically require more bandwidth. “Better compression significantly increases bandwidth, providing new opportunities and leading to economic growth,” concludes Besenbruch.
AISTREAM, AI, video compression, bandwidth supply, CO2 generation, emerging technologies, global compression markets, video on demand sector