Advances in 4D Representation: Geometry, Motion, and Interaction
A new survey paper explores the rapidly advancing field of 4D generation and reconstruction, focusing on how 3D geometry evolves over time with motion and interaction. The paper highlights representative works rather than an exhaustive list, aiming to guide readers in selecting and customizing appropriate 4D representations for their specific tasks. It covers popular methods like NeRFs and 3D Gaussian Splatting, alongside less-explored areas such as structured models and long-range motions, while also discussing the role and limitations of large language models and video foundational models in this domain. AI