Medial Axis Shape Coding in Macaque Inferotemporal Cortex

Abstract

SUMMARY The basic, still unanswered question about visual object representation is this: what specific information is encoded by neural signals? Theorists have long predicted that neurons would encode medial axis or skeletal object shape, yet recent studies reveal instead neural coding of boundary or surface shape. Here, we addressed this theoretical/experimental disconnect, using adaptive shape sampling to demonstrate for the first time explicit coding of medial axis shape in high-level object cortex (macaque monkey inferotemporal cortex or IT). Our metric shape analyses revealed a coding continuum, along which most neurons represent a configuration of both medial axis and surface components. Thus IT response functions embody a rich basis set for simultaneously representing skeletal and external shape of complex objects. This would be especially useful for representing biological shapes, which are often characterized by both complex, articulated skeletal structure and specific surface features.

Bio-inspired Dynamic 3D Discriminative Skeletal Features for Human Action Recognition

“…We first briefly describe the 3D static shape features used by Hung et al [11] to represent the shape space in primate visual cortex. Since we are interested in moving 3D shapes, we represent a moving skeleton as a time-series of these 3D shape features and model the dynamics using a hierarchy of LDSs.…”

“…In Yamane et al [28] and Hung et al [11], a 3D shape is very briefly shown, by using shading and disparity cues, to a macaque monkey with their head restrained so that they cannot move. The electrical activity of several neurons is then recorded.…”

“…In this paper, we propose building 3D representations of human shape and motion that are inspired by recent findings from neuroscience in the work of Yamane et al [28] and Hung et al [11] that attempt to find parametric models for shape space representation in the primate infero-temporal (IT) cortex. We extend their work by proposing representation of human activities as a trajectory through the 3D shape space.…”

“…In Section 2, we provide a brief description of the features proposed by Hung et al [11] to describe a 3D shape space representation in primate cortex. We then briefly outline the background of LDSs and standard metrics used to compare LDSs for classification of time-series data.…”

Image statistics underlying natural texture selectivity of neurons in macaque V4

“…Indeed, previous studies on neural texture representation were typically based on strictly limited samples (6-8)-e.g., 15 different textures were used in Freeman et al (7)-despite a potentially large number of parameters describing the neural selectivity for various natural textures. In this study, we overcome this problem of dimensionality by combining the parametric texture synthesis with an adaptive sampling procedure (23)(24)(25) that efficiently sampled a portion of the stimuli that were expected to evoke stronger responses in the recorded neuron, as well as linear regression with a dimension-reduced version of the PS statistics. As a result, we successfully fitted neural responses to hundreds of textures using PS statistics.…”