Neural Information Processing

BOiS - Berlin Object in Scene Database

In order to understand the human visual system, experimenters often employ “natural stimuli”, typically referring to photographs of the world. However, in order to investigate natural viewing behavior, stimulus material should be created under controlled conditions and be free of optical distortions.

The BOiS - Berlin Object in Scene Database allows the generation of a variety of stimuli for visual search experiments. It comprises photographs of 130 everyday scenes: two with a typical object at two different locations - expected versus unexpected location - and one where the target object is absent, as well as 240 photographs of each of the target objects from various viewpoints. All images were photographed under specified conditions and corrected for photographic artifacts.

The database is enriched with meta-information that allows researchers to generate stimulus sets in which bottom-up features and top-down information psychophysical features can be quantified. We generated experimentally obtained contextual prior maps that quantify people’s expectations to find the target object at a certain location within a scene. To facilitate the extraction of psychophysical features, we provide segmentation masks for the target objects in the scenes. This enables the controlled investigation of bottom-up features and top-down information.