Shengqiong Wu, Meng Luo, Bobo Li, Ziwei Liu, Shuicheng Yan, Ming-Hsuan Yang, Philip Torr, Hao Fei
This survey examines video foundation models and video world models through the lens of world modeling, organizing the literature around four operations: reading, writing, sharing, and interacting. The paper highlights a shift in emphasis from clip-level recognition to persistent state and controllable evolution. It distills open challenges in the field.
Understanding and generating visual sequences has numerous applications in areas like robotics, autonomous vehicles, and human-computer interaction. Advancements in video world models can lead to more realistic and interactive simulations. This research has the potential to improve decision-making in various fields.
Thank you to arXiv for use of its open access interoperability. arxiv.org. aibrief.ai is not affiliated with or endorsed by arXiv.