CoolFace
Datasetpublic

terminusresearch/photo-aesthetics

Photo Aesthetics has moved The maintained dataset is now available at webshart/terminusresearch-photo-aesthetics. The replacement is a fully repackaged Webshart dataset with 30,032 captioned image samples across 371 indexed shards. Its paired JSON indexes include byte offsets, image geometry, and embedded captions for efficient random HTTP range access. The replacement dataset card contains ready-to-use SimpleTuner and Webshart Python examples. This legacy repository's tar… See the full description on the dataset page: https://huggingface.co/datasets/terminusresearch/photo-aesthetics.

sourceHugging Facemitupdated 1mo agoView on Hugging Face
5likes64downloads
Dataset Card

Photo Aesthetics has moved

The maintained dataset is now available at [webshart/terminusresearch-photo-aesthetics](https://huggingface.co/datasets/webshart/terminusresearch-photo-aesthetics).

The replacement is a fully repackaged Webshart dataset with 30,032 captioned image samples across 371 indexed shards. Its paired JSON indexes include byte offsets, image geometry, and embedded captions for efficient random HTTP range access.

The replacement dataset card contains ready-to-use SimpleTuner and Webshart Python examples.

This legacy repository's tar archives and parquet file were removed after the migration completed to avoid storing a second copy of the roughly 397 GB source dataset. Existing users should update their dataset ID to webshart/terminusresearch-photo-aesthetics and use subfolder="webshart".

The original Photo Aesthetics dataset was assembled from Pexels photographs collected in 2023, with captions generated using CogVLM. Migration and Webshart conversion were performed by bghira from Terminus Research.