# YouTube UGC Dataset: User-Generated Content Video Clips with Mean Opinion Scores for Video Quality Assessment and Compression (Google)

## Facts

- Provider: Google / YouTube Media Algorithms
- License: CC-BY-3.0 (https://media.withyoutube.com/)
- Stored at: Google Cloud Storage, gs://ugc-dataset (us)
- Access: public, readable without an account
- Category: Computer Vision
- Source: https://github.com/google/uvq
- Page: https://scigantic.com/archive/743e8d1c-ecec-5a74-b9f1-a70c71232427
- Keywords: Computer Vision, benchmark, Google Cloud Storage

## About

The YouTube UGC dataset is a sample of about 1,500 user-generated content (UGC) video clips, 20 seconds each, taken from videos uploaded to YouTube under a Creative Commons license and chosen for real perceptual quality problems such as blockiness, blur, banding, noise and jerkiness. Google released it for research on video compression and perceptual video quality assessment where the original upload is not pristine. Categories are Animation, Cover Song, Gaming, HDR, How-To, Lecture, Live Music, Lyric Video, Music Video, News Clip, Sports, Television Clip, Vertical Video, Vlog and VR, at 360p to 1080p and 4K for some genres. Every clip has crowdsourced mean opinion scores (MOS, 1 to 5, 100+ raters per clip), with extra MOS for three overlapping 10 second chunks, differential MOS (DMOS) for Gaming, Sports and Vlog, and content labels. The bucket holds original_videos/ (clips by category plus the MOS, DMOS and content label spreadsheets), original_videos_h264/ (H.264 transcodes), vp9_compressed_videos/ (VP9-compressed versions), previews/, the dataset and CoINVQ papers, a LICENSE file and an ATTRIBUTION file naming each video's creator. Use it to train or benchmark no-reference video quality (VQA) models such as UVQ and to evaluate codecs on real-world content; read with ffmpeg, PyAV or OpenCV. License: Creative Commons Attribution (CC BY 3.0) per the dataset page and LICENSE file; credit the creators in ATTRIBUTION. Cite Wang, Inguva and Adsumilli, YouTube UGC dataset for video compression research (IEEE MMSP 2019).

## Load the data

```python
import gcsfs
fs = gcsfs.GCSFileSystem(token='anon')
fs.ls('ugc-dataset/', detail=False)[:10]
```

Lists the files with anonymous access. Runs on any machine, with no cloud account.

## Data structure

Example file: `original_videos/Animation/1080P/Animation_1080P-05f8.mkv`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/743e8d1c-ecec-5a74-b9f1-a70c71232427
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/743e8d1c-ecec-5a74-b9f1-a70c71232427
- Over MCP: `get_schema_card` with id `743e8d1c-ecec-5a74-b9f1-a70c71232427` at https://api.scigantic.com/mcp

## Related topics

- [Benchmark datasets](https://scigantic.com/topics/benchmark)
- [Computer Vision datasets](https://scigantic.com/topics/computer-vision)
