# CommonCatalog CC-BY-SA: Flickr ShareAlike images with synthetic captions

## Facts

- Provider: MosaicML / Databricks (Aaron Gokaslan et al.)
- License: CC-BY-SA-4.0 (https://huggingface.co/datasets/common-canvas/commoncatalog-cc-by-sa)
- Size: 4,131 files
- Format: Parquet (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/common-canvas/commoncatalog-cc-by-sa
- Access: public, readable without an account
- Category: Computer Vision
- Source: https://arxiv.org/abs/2310.16825
- Page: https://scigantic.com/archive/a83d871f-51c9-591d-8bda-533698bda6a4
- Keywords: Computer Vision, Parquet, Hugging Face

## About

This is the CC BY-SA partition of CommonCatalog, the openly licensed image-caption collection MosaicML built to train the CommonCanvas latent diffusion models (Gokaslan et al., arXiv:2310.16825). Like the CC BY partition, it draws on Yahoo Flickr Creative Commons 100M (YFCC100M) photographs uploaded by Flickr users around 2014, keeps only images under a ShareAlike Creative Commons licence that allows commercial use, and pairs each with a synthetic BLIP-2 caption. Parquet rows contain the JPEG bytes, the BLIP-2 caption, the Flickr title, user and machine tags, capture device, date taken, geolocation where present, original and stored image dimensions, Flickr page and download URLs, and the per-photo licensename and licenseurl, so derived works can carry correct attribution and share-alike notices. Images go up to 4K resolution. Measured 2026-11-14: ten top-level subsets (0 to 9), each partitioned by least_dim_range and aspect_ratio_bucket folders of Parquet files; the Hub reports about 13.6 TB of storage. Read with pyarrow, pandas or the datasets library in streaming mode. Uses: text-to-image diffusion training, image captioning, provenance and attribution studies. Every image is under a CC BY-SA licence from its Flickr author; attribute the photographer and share adaptations alike.

## Load the data

```python
from datasets import load_dataset
ds = load_dataset("common-canvas/commoncatalog-cc-by-sa")
print(ds)
```

## Data structure

File types in sampled files: Parquet

Example file: `0/least_dim_range=1024-2048/aspect_ratio_bucket=1-1/part-00000-tid-7761264679195244086-4c6ec9dc-fb3d-4bd3-98ec-f987466013dd-3488916-1-c000.parquet`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/a83d871f-51c9-591d-8bda-533698bda6a4
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/a83d871f-51c9-591d-8bda-533698bda6a4
- Over MCP: `get_schema_card` with id `a83d871f-51c9-591d-8bda-533698bda6a4` at https://api.scigantic.com/mcp

## Related topics

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