Datasets:

FAU-LMS
/

UGC360

license: cc-by-4.0
tags:
  - 360-degree
  - video compression
  - spherical video
  - image sequence
  - video
  - pytorch-compatible
pretty_name: UGC360
size_categories:
  - 10K<n<100K
configs:
  - config_name: UGC360-S
  - config_name: UGC360-M
  - config_name: UGC360-L

UGC360

Dataset of 6866 9-frame 360-degree video sequences collected from 1321 unique videos from Vimeo and Youtube licensed under a Creative Commons License. The dataset is split into three subsets UGC360-S, UGC360-M, and UGC360-L to ease data handling.

Subset	Resolutions	Minimum	Maximum	Clips
UGC360-S Part 1, UGC360-S Part 2, UGC360-S Part 3, UGC360-S Part 4	.5K - 2.5K	640x320	2560x1440	5080
UGC360-M Part 1, UGC360-M Part 2, UGC360-M Part 3	3K - 6K	3072x1536	5760x2880	1618
UGC360-L	8K	7680x3840	7680x4320	168

Unpacking the dataset

Due to file size limitations on the Hugging Face platform, we had to upload the UGC360-S and UGC360-M subset archives as multipart zip files. To unzip these multipart zips, download all multipart zip files (.z01, z02, ..., .zip) of a subset to the same folder. From that folder, execute

zip -s- UGC360-S.zip -O UGC360-S-all.zip
unzip UGC360-S-all.zip

Replace 'UGC360-S' by 'UGC360-M' to do the same for the UGC360-M subset. The directory structure of the final dataset should then follow

.
├── UGC360-L
│   ├── ugc360-l.csv
│   ├── <video_id>/
│   │   ├── <clip_id>/
│   │   │   ├── 01.png
│   │   │   ├── 02.png
│   │   │   ├── 03.png
│   │   │   ├── ...
│   │   │   └── 09.png
│
├── UGC360-M
│   ├── ugc360-m.csv
│   ├── <video_id>/
│   │   ├── <clip_id>/
│   │   │   ├── 01.png
│   │   │   └── ...
│
└── UGC360-S
    ├── ugc360-s.csv
    ├── <video_id>/
    │   ├── <clip_id>/
    │   │   ├── 01.png
    │   │   └── ...

License

The UGC360 subset is licensed under Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0).

Each individual clip within the dataset originates from videos published under various Creative Commons licenses (e.g., CC-BY, CC-BY-SA). The corresponding license for each clip is documented in the CSV metadata files within each subset folder:

UGC360-S/ugc360-s.csv
UGC360-M/ugc360-m.csv
UGC360-L/ugc360-l.csv

Each row contains a license column indicating the license of the source video.

Attribution and Compliance

We collected only videos explicitly marked as Creative Commons on their original platforms (YouTube or Vimeo). For each clip, the dataset includes attribution information:

Video ID
Clip ID
Publisher
Original video URL
License type
Publication date

If you use this dataset, you are required to:

Comply with the CC license of each video
Attribute the dataset as a whole as:

UGC360 Dataset (CC BY-SA 4.0)

If you redistribute or modify this dataset, you must do so under the same license (CC BY-SA 4.0).

Publication

If you use this dataset, please cite

@inproceedings{regensky2025nvc360,
  title     = {Beyond Perspective: Neural 360-Degree Video Compression},
  author    = {Andy Regensky and Marc Windsheimer and Fabian Brand and André Kaup},
  booktitle = {accepted for the IEEE/CVF International Conference on Computer Vision (ICCV)},
  year      = {2025}
}