NUS-WIDE stands for a real-world web image database from National University of Singapore. It is a multi-label web image dataset created by the Media Search Lab of the National University of Singapore. This dataset can promote the performance of existing image annotation and retrieval methods.

The dataset includes:

• 269,648 images, 5,018 specific tags
• 6 types of low-level features, including 64-D color histogram, 144-D color correlation map, 73-D edge histogram, 128-D microwave texture, 225-D block color moment of 5×5 fixed grid partition and 500D bag of words based on SIFT description
• 81 concepts, including goals and scenarios

In addition, this dataset also provides:

1. A lightweight NUS-WIDE dataset, named NUS-WIDE-LITE
2. Object image dataset, named NUS-WIDE-OBJECT
3. Scene image dataset, named NUS-WIDE-SCENE

Example image: