News
2019-07-12 An Artwork Photo dataset are added, thanks for the efforts from FongXue Chen, BingYu Shen, and XiaoWen Li. This dataset can be used to investigate the performance decrease of artwork retrieval, classification leading by different issues (e.g., rotation, blur, etc.) in the artwork photos.
2019-05-05 A toy artwork dataset with 43,455 images are added.
2018-12-03 Some new statistic results on the dataset are added.
Introduction
ART500K is a large-scale visual arts dataset with more than 500K images, each with over 10 attribute labels, apart from some general labels (e.g., artist, genre, art movement), some special labels (e.g., event, historical figure, description) are included. The dataset can be used in different tasks (e.g., visual arts classification, viusal arts retrieval, visual arts image caption, etc.). Both computer science community and visual arts community can get benefits from the dataset.
Downloads
- Raw images of visual arts with general label list: Data&Labels(56.6GB)/Labels
- Raw images of visual arts with event label list: Data(15GB)/Labels
- Raw images of visual arts with historical figure label list: Data(26GB)/Labels
- Raw images of visual arts with place label list: Place1(86GB)/Place2(85GB)/Place3(73GB)/Labels
- Toy artwork dataset with 43,455 images: Data (7GB)/Labels
- Artwork photos: Data (62GB)/Introduction
Agreement and Disclaimer
- The ART500K dataset is available for non-commercial research purposes only.
- All images of the ART500K dataset are obtained from the Internet which are not property of HKUST-NIE Social Media Lab, The Hong Kong University of Science and Technology. The Lab is not responsible for the content nor the meaning of these images.
- You agree not to reproduce, duplicate, copy, sell, trade, resell or exploit for any commercial purposes, any portion of the images and any portion of derived data.
- You agree not to further copy, publish or distribute any portion of the ART500K dataset. Except, for internal use at a single site within the same organization it is allowed to make copies of the dataset.
- The Lab reserves the right to terminate your access to the ART500K dataset at any time.
Citation
@inproceedings{mao2017deepart,
title={Deepart: Learning joint representations of visual arts},
author={Mao, Hui and Cheung, Ming and She, James},
booktitle={Proceedings of the 25th ACM international conference on Multimedia},
pages={1183--1191},
year={2017},
organization={ACM}
}
@article{mao2019visual,
title={Visual Arts Search on Mobile Devices},
author={Mao, Hui and She, James and Cheung, Ming},
journal={ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM)},
volume={15},
number={2s},
pages={60},
year={2019},
publisher={ACM}
}
Related Work
The following paper employed ART500K for visual arts feature learning.H. Mao, M. Cheung, and J. She, "DeepArt: Learning Joint Representations of Visual Arts", in ACM International Conference on Multimedia (MM), 2017