About this dataset

Description

This dataset contains over 1600 paintings from the Russian Museum, the Tretyakov Gallery and other public archives. Each painting has a corresponding description generated by GPT2, which can be found in description.csv. The dataset was created to finetune Stable Diffusion and other text-to-image models. All of the text data is available in both English and Russian languages.

Dataset Structure

description.csv: the file containing the annotations

data: the folder containing the paintings in .jpg format

Acknowledgements

https://gallerix.ru/

https://rusmuseumvrm.ru/