Developmental Stereo: Topographic Iconic-Abstract Map from Top-Down Connection

Abstract

Abstract. Engineering approaches to stereo typically use explicit search for the best matching between left and right sub-windows, which involves a high cost of search and unstable performance in the presence of binocular inconsistency and weak texture. The brain does not seem to conduct explicit search in the V1 and V2 cortex. But the mechanisms that the brain employs to integrate binocular disparity into 3-D perception is still largely a mystery. The work presented in this paper focuses on an important issue of integrated stereo: How the same cortex can perform recognition and perception by generating a topographic disparity-tuning map using top-down connections. As top-down connections with objectclass supervisory signals result in topographic class maps, the model presented here clarifies that stereo can be processed by a unified in-place learning framework in the neural layers, and can generate iconic-abstract internal representation.

Developmental Stereo: Emergence of Disparity Preference in Models of the Visual Cortex

“…They reported a 90% correct recognition rate for 11 classes of disparity. In Solgi and Weng 2008 [28], a multilayer in-place learning network was used to detect binocular disparities that were discretized into classes of 4 pixels intervals from image rows of 20 pixels wide. This classification scheme does not fit well for higher accuracy needs, as a misclassification between disparity class and class 0 is very different from that between a class and class 4.…”

“…Similar to their work, in this paper the emergent recursive topdown context is provided from the response pattern of the motor cortex at the previous time to the feature detection cortex at the current time. Biologically plausible networks (using Hebbian learning instead of error back-propagation) that use both bottom-up and top-down inputs with engineering-grade performance evaluation have not been studied until recently [15], [28], [33].…”

Temporal information as top-down context in binocular disparity detection

“…Similar to their work, in this paper the emergent recursive top-down context is provided from the response pattern of the motor cortex at the previous time to the feature detection cortex at the current time. Biologically plausible networks (Hebbian learning instead of error back-propagation) that use both bottom-up and top-down inputs with engineering-grade performance evaluation have not been studied until recently [13], [14].…”

Early Perception-Action Cycles in Binocular Vision

“…Similarly, in the ICT community there are several examples of neuromorphic (i.e., distributed) approaches profitably used to challenge conventional solutions to computer vision problems, by introducing sophisticated interpretation of biologically plausible operations (e.g., Tsang and Shi (2007), Tsang and Shi (2009), Bayerl and Neumann (2007); see also Franz and Triesh (2007), Wang and Shi (2009), Solgi and Weng (2008)).…”