Information about the data:
The Object images were rendered from 3D planes. Objects vary in five dimensions: size, position (x, y), rotation in-depth, and rotation in plane. There are five object categories with, on average, 16 3D planes for each category. To alter the complexity of the images, we generated images across seven levels of variations: starting from zero level variation, in which no variation was applied to 3D object planes, to seventh level of variation, in which substantial variations were applied to images. In each level, we randomly sampled different values for each dimension (e.g., size, rotation, and position) from a uniform distribution and finally the selected values were applied to a 3D plane.
please enter your information below; then, the database will be available to download.
Please cite the following publication when using the dataset.
Ghodrati M, Farzmahdi A, Rajaei K, Ebrahimpour R, Khaligh-Razavi S-M (July 2014), Feedforward Object-Vision Models Only TolerateSmall Image Variations Compared to Human. Front. Comput. Neurosci. | doi: 10.3389/fncom.2014.00074.