一个视频,一个世界:将单目视频转化为物理4D场景
TLDR
OVOW无需训练,从单目视频重建实例级可仿真4D网格场景,采用四阶段流水线。
评分理由
The paper presents a novel training-free pipeline for 4D reconstruction with instance separation and physical plausibility, which is a strength. However, evaluation is limited to synthetic benchmarks, and the core contribution is about reconstruction rather than world models, making relevance to the given keywords low.
Read-first 评分解释
综合优先阅读分 23.4,由主题、引用、图谱、方法、可复现性和近期性等信号加权得到。 原始总分保留为 4。
研究版图角色
前沿论文
排序敏感性
稳定性:volatile;排名波动范围:43。