If you use this dataset, please cite the following paper in which it has been first introduced:
Visual Word Disambiguation by Semantic Contexts,
Yu Su and Frederic Jurie,
ICCV, 2011. [pdf]
The QUAERO still image dataset contains 518 visual concepts, with about 1000 images per concept. The concepts in this dataset are very diverse, including animal, plant, scene, object, event, people and some general concepts. For each concept, at most 1000 images were downloaded by querying Google image search. Images in the dataset were manually annotated as relevant, irrelevant or ambiguous to their corresponding concepts.
This dataset is the property of the University of Caen Basse-Normandie