In biologically motivated models for object segmentation and recognition, the ability to do stereo vision is often overlooked. Here I will present a model inspired by the ventral stream of the human visual cortex, which integrates disparity information of local features into a pattern recognition hierarchy.
The model will show that depth as a feature can be learned from disparity input and how disparity features can help us interpret certain corner cases of object shape. I will also show how disparity information may be bound to visual pattern information and how this relates to a possible, efficient solution of the binding problem between object information and position.
There will also be a real-time presentation of a partial model implementation.