What do you want to download?

Select the data you wish to download below.

Paintings

Bounding boxes

All paintings

All the ~19k paintings in the dataset

Only a selection of Paintings

For example, only those that depict fabrics, or only those by a specific author, or ..

Boxes with padding

In this option, all bounding boxes add padding around the edges of the material

Boxes without padding

In this option, all bounding boxes are cut execatly at the edges of the material

All boxes

The boxes manually annotated by participants, as well as the automatic bounding boxes

Only a selection of the boxes

For example only the boxes that contain a specific material, those that were made in a certain time-period or only those that were made by human annotators

All human boxes

All the bounding boxes created by human annotations

All automatic boxes

All the bounding boxes created by a neural network

Only a selection

Take a selection of the database, for example only those boxes that contain a specific material or that was created at a specific time-period

How to download:

The easiest way to download all the paintings is by downloading them directly from the 4TU datarepository.

All paintings.

All paintings metadata.

There are three steps to download only a selection of the paintings.

First, use the search tool on the home page to make your selection.
Once you have made a selection, a download button will appear. This will download all the meta data in an excel file. Note that this excel file is generated specifically for your selection, so this might take a few seconds if you have a large selection.
Last, you can use the Python script below to download the images. If you have a large selection, it will likely be faster to download all paintings and then select those you want.

Click here to download the Python file!

The easiest way to download all the bounding boxes (with padding) is by downloading these two files directly from the 4TU datarepository.

The images:

All boxes with padding created by humans
All boxes with padding created by a neural network

The metadata:

Metadata for boxes with padding created by humans
Metadata for boxes with padding created by a neural network

The easiest way to download all the bounding boxes (without padding) is by downloading these two files directly from the 4TU datarepository.

human_bbox_1024.zip , all boxes without padding created by humans
automatic_bbox_1024.zip , all boxes without padding created by a neural network

The metadata:

Metadata for boxes with padding created by humans
Metadata for boxes with padding created by a neural network

The easiest way to download all the human made bounding boxes (with padding) is by downloading them directly from the 4TU datarepository.

All boxes with padding created by humans
Metadata for all boxes created by humans

The easiest way to download all the automatic bounding boxes (with padding) is by downloading them directly from the 4TU datarepository.

All boxes with padding created by a neural network
Metadata for all automatically created by humans

The easiest way to download all the human made bounding boxes (with padding) is by downloading them directly from the 4TU datarepository.

human_bbox.zip , all boxes without padding created by humans
Metadata for all boxes created by humans

The easiest way to download all the automatic bounding boxes (with padding) is by downloading them directly from the 4TU datarepository.

automatic_bbox_1024.zip , all boxes without padding created by a neural network
Metadata for all automatically created by humans

There are three steps to download only a selection of bounding boxes with padding.

First, use the search tool on the home page to make your selection
Once you have made a selection, a download button will appear. This will download all the meta data in an excel file. Note that this excel file is generated specifically for your selection, so this might take quite long if you have a large selection.
Last, you can use the Python script below to download the images. If you have a large selection, it will likely be faster to download all boxes and then select those you want.

Click here to download the Python file!

There are three steps to download only a selection of bounding boxes without padding.

First, use the search tool on the home page to make your selection
Once you have made a selection, a download button will appear. This will download all the meta data in an excel file. Note that this excel file is generated specifically for your selection, so this might take quite long if you have a large selection.
Last, you can use the Python script below to download the images. If you have a large selection, it will likely be faster to download all boxes and then select those you want.

Click here to download the Python file!

The downloads (paintings and bounding boxes) are all at a max resolution of 1024x2014. Need a higher resolution?

You can download all paintings at the maximum resolution that we have through this link: Click here!

The bounding boxes are available at a max resolution of 1024x1024. These are taken from the max resolution paintings, and then resized to fit a 1024x1024 box (only if the bounding box would be bigger than the max resolution). If you wish to have bounding boxes above this resolution, you will have to download the paintings at the max resolution and use the data in the .csv files to create the boxes at the desired size. You can create the .csv files through making a selection in the gallery, or by downloading this .csv that contains all the bounding boxes: Click here!