Abstract
Multidimensional 3D-rendered objects are an important component of vision research and video-gaming applications, but it has remained challenging to parametrically control and efficiently generate those objects. Here, we describe a toolbox for controlling and efficiently generating 3D rendered objects composed of ten separate visual feature dimensions that can be fine-adjusted using python scripts. The toolbox defines objects as multi-dimensional feature vectors with primary dimensions (object body related features), secondary dimensions (head related features) and accessory dimensions (including arms, ears, or beaks). The toolkit interfaces with the freely available Blender software to create objects. The toolbox allows to gradually morph features of multiple feature dimensions, determine the desired feature similarity among objects, and automatize the generation of multiple objects in 3D object and 2D image formats. We document the use of multidimensional objects in a sequence learning task that embeds objects in a 3D-rendered augmented reality environment controlled by the gaming engine unity. Taken together, the toolbox enables the efficient generation of multidimensional objects with fine control of low-level features and higher-level object similarity useful for visual cognitive research and immersive visual environments.