3D Vision

Three-dimensional vision has been revolutionized by Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS), enabling photorealistic novel view synthesis and real-time rendering. Large Language Models have stepped into the 3D world, unlocking open-vocabulary scene understanding, spatial reasoning, and language-guided 3D generation. The convergence of neural scene representations with simultaneous localization and mapping (SLAM) has catalyzed a paradigm shift — from NeRF-based dense mapping to real-time Gaussian Splatting SLAM. Combined with classical multi-view stereo, structure-from-motion, and robotics applications, these advances form a rich ecosystem spanning from data capture to interactive 3D content creation and autonomous navigation. Explore the six sub-domains below.