3D Geometry

3D geometry research focuses on accurately and efficiently representing and manipulating three-dimensional shapes from various input sources, such as single images or point clouds. Current efforts concentrate on developing novel neural network architectures, including those based on neural radiance fields (NeRFs), signed distance functions (SDFs), and Gaussian splatting, to improve the accuracy and realism of 3D reconstructions, often incorporating techniques like active pattern projection and test-time adaptation to handle challenging scenarios. These advancements have significant implications for diverse fields, ranging from robotics and computer vision (e.g., object removal, pose estimation) to cultural heritage preservation and materials science (e.g., molecular geometry prediction). The development of efficient algorithms for querying and manipulating these neural representations is also a key area of ongoing research.