Sekai: A Video Dataset towards World Exploration
Zhen Li, Chuanhao Li, Xiaofeng Mao, Shaoheng Lin, Ming Li, Shitian Zhao, Zhaopan Xu, Xinyue Li, Yukang Feng, Jianwen Sun, Zizhen Li, Fanrui Zhang, Jiaxin Ai, Zhixiang Wang, Yuwei Wu, Tong He, Jiangmiao Pang, Yu Qiao, Yunde Jia, Kaipeng Zhang
2025-06-19
Summary
This paper talks about Sekai, a large video dataset collected from around the world that includes detailed notes and labels to help AI models learn about exploring different places.
What's the problem?
The problem is that existing video datasets often lack the variety, detail, or global coverage needed for AI to understand and generate videos related to exploring diverse real-world environments.
What's the solution?
The researchers created Sekai, which gathers videos from many parts of the world and includes comprehensive annotations that provide important context and information, helping AI models learn better and improve their ability to handle world exploration tasks.
Why it matters?
This matters because having a rich and well-labeled video dataset helps improve AI’s capability to generate and understand videos about different places, which can be useful for applications like virtual travel, education, and helping robots explore the world.
Abstract
Sekai, a worldwide video dataset with comprehensive annotations, is introduced to support world exploration applications, enhancing video generation models.