metadata

license: mit
tags:
  - photographs
  - photos
  - image-data
  - text-to-image
  - image-caption pairs
  - simpletuner
pretty_name: Photo Anatomy
configs:
  - config_name: default
    data_files:
      - split: train
        path: train.parquet

Photo Anatomy Dataset

Pulled from Pexels in 2023.

Images contain a majority of images of "people holding things".

Image filenames may be used as captions, or, the parquet table contains the same values.

This dataset contains the full images.

Captions were created with CogVLM.