The Oxford pets dataset has been added, which is available as:

• image classification

• image segmentation

• instance segmentation

• object detection (bbox; head or animal).

The CamVid-12 image segmentation dataset has been added. It is a subset of the original CamVid dataset and only has 12 classes instead of 32.

The CamVid image segmentation dataset has been added. It consists of 700 images (with associated annotations) and has 32 distinct labels.

The following flower datasets from the Visual Geometry Group at Oxford University have been added:

• 17 Flowers

• 102 Flowers

The following datasets have been added:

• American Sign Language Letters

• Caltech-UCSD Birds-200-2011

• Stanford Dogs

Datasets used by the UFDL project are now available through this website. As additional datasets become available, they will get posted here.