Visual Computing Group: An Exciting New Chapter
The Visual Computing Group (VCG) was formed in 2024 from the merger of the iSIZE team with parts of the SIE Game Platform Artificial Intelligence (GPAI) group. This created a large team of R&D engineers with unique technical skills in: neural-networks, video coding, quality assessment, generative models, game rendering, and high-performance computing. The group’s mission is to design and deploy advanced neural networks and machine learning (ML) for game rendering and streaming systems of SIE that exceed the state-of-the-art in runtime efficiency, visual quality and latency.
Recent examples of our work include:
- the first approach for perceptual video compression with neural wrapping, improving bitrate efficiency of state-of-the-art codecs by around 20%;
- state-of-the-art no-reference metrics for image quality assessment;
- foundation models for perceptual similarity assessment;
- generative models for photorealistic neural avatars that avoid uncanny valley problems;
- novel ways to combine streaming and VR rendering or neural rendering and streaming;
- extensive outreach to academia and research collaboration programmes with top-tier Universities around the world.
These are only glimpses of the work we do internally in support of future product design and prototyping within SIE.
And our impact is growing. We have recently launched a new initiative within SIE that will explore rendering and streaming in completely novel ways. This aims to go significantly beyond all current work on ML-based game rendering, which may only be scratching the surface of what is possible when game rendering, AI and distributed computing come together. To support this new initiative and strengthen our team’s efforts in this direction, we have opened multiple R&D positions in our group.
If building the future of gaming based on computer vision, gen-AI, and game rendering excites you, please consider joining us! We are currently looking to hire for the below roles:
Related Posts
