OpenAccess - The Met Collection
This dataset brings together more than 470,000 works from the collections of the Metropolitan Museum of Art, along with their comprehensive metadata and, in many cases, high-resolution images that are completely royalty-free.
470,000+ works, tabular format (CSV/parquet), metadata + associated visuals
CC0 1.0
Description
The dataset OpenAccess — The Met Collection contains detailed information on more than 470,000 works preserved at the Metropolitan Museum of Art in New York. For each work, there is metadata such as title, artist, date, medium, department, and a link to the freely accessible high-resolution image.
What is this dataset for?
- Train computer vision models (classification, image search, artistic OCR...)
- Create educational or cultural applications based on high quality artistic images
- Explore or enrich knowledge bases related to art history and world heritage
Can it be enriched or improved?
Yes, it is possible to annotate images for specific uses (artistic style, period, iconography), or to cross this dataset with other open cultural databases (Wikidata, Europeana...). You can also generate automatic descriptions or visual summaries for each work via LLM + vision models.
🔎 In summary
🧠 Recommended for
- Museum projects
- Cultural visual models
- Education and artistic visualization
🔧 Compatible tools
- TensorFlow
- PyTorch
- CLIP
- OpenCV
- Streamlit
- Label Studio
💡 Tip
Filter works with high-resolution images to create custom games tailored to machine learning.
Frequently Asked Questions
Are all images in the dataset royalty-free?
Yes, all images included in the dataset are under the CC0 license, freely reusable even for commercial purposes.
Can this dataset be used for digital educational or artistic projects?
Absolutely, this dataset is ideal for educational initiatives, digital exhibitions or cultural applications.
What is the origin of the works included?
They come from the collections of the Metropolitan Museum of Art in New York, covering all regions of the world and several millennia.




