The role of independent motion in object segmentation in the ventral visual stream: Learning to recognise the separate parts of the body

This paper investigates how the visual areas of the brain may learn to segment the bodies of humans and other animals into separate parts. A neural network model of the ventral visual pathway, VisNet, was used to study this problem. In particular, the current work investigates whether independent motion of body parts can be sufficient to enable the visual system to learn separate representations of them even when the body parts are never seen in isolation. The network was shown to be able to separate out the independently moving body parts because the independent motion created statistical decoupling between them.

Research highlights
► The role of independent motion in visual object segmentation into parts is examined.
► Biologically plausible neural network model of the ventral visual stream, VisNet, is used.
► Explains underlying brain mechanisms for geometrical theories of visual object segmentation.
► Lack of independent motion results in one representation of the whole object formed.
► Independent motion results in separate representations of the full body and its constituent parts.