Session

DescriptionThis session is about 3D video processing, diffusion, Gaussian splatting, and their memory optimizations. As research into deep learning grows, often people have found that the models used in deep learning are transitive to applications in vision computing, thus creating a cross-disciplinary experience. We cover hardware-efficient diffusion acceleration, 3D video generation models, and 3D Gaussian splatting. After our 3 papers that focus on Gaussian splatting, we conclude with a paper about diffractive optical neural network, using an emergent technology technique to attack the issues of speed and energy efficiency.