LagerNVS: Latent Geometry for Fully Neural Real-time Novel View Synthesis
Researchers have developed LagerNVS, a novel neural network designed for real-time novel view synthesis. This system leverages "3D-aware" latent features, with its encoder pre-trained on explicit 3D supervision. LagerNVS achieves state-of-the-art performance, demonstrating high PSNR scores and real-time rendering capabilities, while also showing generalization to in-the-wild data. AI
IMPACT This research introduces a new method for generating novel views of 3D scenes in real-time, potentially impacting applications in computer graphics and virtual reality.