📊A huge dataset of images and their captions
Pixel Prose is a dataset that contains over 16 million diverse images from three different web databases (commonPool, CC12M, RedCaps) with captions created using Google Gemini 1.0 Pro Vision.
The following Python script can be used to load a dataset using the API:
from datasets import load_dataset
# for downloading the whole data
ds = load_dataset("tomg-group-umd/pixelprose")
Post #693
1.5K
