From NVIDIA: 2024-12-09 09:00:58

NVIDIA Research and the University of Maryland have developed QUEEN, an AI model that enables free-viewpoint video streaming, allowing viewers to experience a 3D scene from any angle. QUEEN will be presented at the NeurIPS conference, showcasing its ability to balance compression rate, visual quality, and rendering time for optimal streaming.

QUEEN’s innovative approach to free-viewpoint video streaming offers high-quality visuals even in dynamic scenes, rendering faster than previous methods. By tracking and reusing static elements in a scene, QUEEN achieves impressive performance benchmarks on NVIDIA Tensor Core GPUs, enabling immersive experiences in media broadcasts and warehouse operations.

The release of QUEEN’s code as open source will facilitate its application in various fields, including media broadcasts, robotics, and video conferencing. NVIDIA’s extensive research at NeurIPS includes over 50 groundbreaking AI projects with potential applications in simulation, robotics, and healthcare, highlighting the company’s commitment to advancing AI technologies.



Read more at NVIDIA: NVIDIA Research Model Enables Dynamic Scene Reconstruction