• 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)/Label
• Toy artwork dataset with 43,455 images: Data (7GB)/Labels
• Artwork photos: Data (62GB)/Introduction

• train.zip -zip file containing the images in the training set (.jpg)
• train_{1,2,..9}.zip – subsets of train.zip since the data size is large. You don’t need any of these if you can download train.zip.
• test.zip – zip file containing the images in the test set (.jpg)
• train_info.csv – file listing the image filename, artistID, genre, style, date and title for images in the training set.
• submission_info.csv – each row lists an index and the filenames of the two images for which the algorithm needs to make a prediction about whether they were created by the same artist.
• sampleSubmission.csv – a sample submission file with a column for comparison number and the predicted probability
• replacements_for_corrupted_files.zip – 10 files in train and test sets that were corrupted images, these are the correct images

BHL provides data exports and APIs to allow individual users and data providers to download, remix and reuse BHL content.

Exports of BHL bibliographic, scientific name, and full optical character recognized text are available in a variety of formats.
A series of files is available for download that will enable libraries and other data providers to identify digitized titles available within BHL. These files include metadata about each volume scanned, as well as information about the millions of scientific names that have been identified throughout the BHL corpus and the pages on which those names occur.

In The Art Genome Project, every gene (category) is applied along with a value from 1-100 that indicates its relevance to the particular work, artist, designer or architect.

However, this list may still be used as subject headings, tags, or general categories for organizing collections of artworks, artists, designers, functional objects, and more.

The Prints & Photographs Online Catalog includes images in the following formats:

• gif – generally small “thumbnails” used for previewing images; a gif image displays at the top of its associated catalog record and, in some cases, it is the only image that will display to those searching outside the Library of Congress because of rights considerations (extension on the file name is .gif). The resolution is generally about 150×150 pixels.
• jpeg – generally a larger image that displays in a separate screen from the catalog record; sometimes two types of jpeg files are available–one for reference viewing and one at a higher resolution (extension on the file name is .jpg).
• tiff- generally the highest resolution file available in PPOC, viewed or downloaded via a link on the screen where the reference jpeg displays (extension on the file name is .tif).

The database includes the recognition status (i.e. whether the model recognized the modifier, indicated with “Yes”, “No” or “Unsure”) of each individual modifier for each set of models. It also includes examples of images synthesized by the above models (Figure 1) when tested with the specified modifier to determine recognition status (see FAQ below – “How do you determine whether the models recognize a modifier?”).

The dataset provides data records relating to the 130,000+ artworks in our collection and the artists who created them. You can download the dataset free of charge without seeking authorization from the National Gallery of Art.
The dataset is published in CSV format and uses UTF-8 encoding, and is updated daily. Links and references to images and other media such as audio and video files are contained in the dataset, but the images and media files themselves are not included under this program.