Skip to content

Latest commit

 

History

History
57 lines (48 loc) · 3.87 KB

File metadata and controls

57 lines (48 loc) · 3.87 KB

360World Dataset

A Curated Collection of 360° Video Clips for Research Use

Data Availability Statement

The 360World dataset is built exclusively from publicly accessible YouTube content; no proprietary or private sources are used. Instead of redistributing raw videos, we release only the YouTube video IDs required for download. This release is intended strictly for non-commercial research purposes and adheres to YouTube’s privacy and fair-use policies. No user-identifiable information is collected or disclosed.

Clarifications

  • Copyright of each video remains with the original uploader.
  • All content is sourced from the public Internet; our institutions claim no ownership and assume no responsibility for its meaning or legality.
  • By using this dataset, you agree not to reproduce, duplicate, sell, trade, or otherwise exploit any portion of the videos or derived data for commercial purposes. Further redistribution, publication, or mass distribution of the dataset is prohibited.

Dataset Construction Pipeline

1. Data Collection & Filtering

  • Source: Public YouTube videos.
  • Search keywords: “panorama”, “VR”, and related 360° terms.
  • Filtering: Manual review to ensure relevance, high resolution, and absence of policy violations.

2. Processing Workflow

  • Shot Segmentation: Key frames are detected via inter-frame difference, after which the video is segmented into clips.
  • Text Annotation: Each clip is captioned with ShareGPT4Video.
  • Text Refinement: Captions are summarized and polished by GPT-4.

Dataset Overview

The 360World dataset consists of 7314 video clips curated from online 360 videos. Each entry in dataset/360video_data.csv contains:

Column Description
YouTube_ID 11-character YouTube video ID. Full URL: https://www.youtube.com/watch?v={id}
Video_ID Composite identifier: {YouTube_ID}_{clip_index}
Caption Concise textual summary of the clip scene.
Frame_Start First frame index (0-based) in the original video.
Frame_End Last frame index (inclusive) in the original video.
FPS Frame rate of the source video.
Total_Duration Duration (s) of the source video.
Total_Frames Total frame count of the source video.

Example Entry

YouTube_ID Video_ID Caption Frame_Start Frame_End FPS Total_Duration (s) Total_Frames
oovURIu4040 oovURIu4040_7 “A serene evening at an outdoor amphitheater. Attendees enjoy the sunset; shadows gradually lengthen.” 3831 3987 30 258.4 7752

License

360World is released under the Creative Commons Attribution 4.0 International License (CC BY 4.0).
When using this dataset, please cite:

@misc{zhou2025holotimetamingvideodiffusion,
      title={HoloTime: Taming Video Diffusion Models for Panoramic 4D Scene Generation}, 
      author={Haiyang Zhou and Wangbo Yu and Jiawen Guan and Xinhua Cheng and Yonghong Tian and Li Yuan},
      year={2025},
      eprint={2504.21650},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2504.21650}, 
}