Rotationally invariant vision recognition with neuromorphic transformation and learning networks
Richard James Sofatzis, Saeed Afshar, Tara Julia Hamilton · 2014
In this paper we present a biologically inspired rotationally-invariant end-to-end recognition system demonstrated in hardware with a bitmap camera and a Field Programmable Gate Array (FPGA). The system integrates the Ripple Pond Network (RPN), a neural network that performs image transformation from two dimensions to one dimensional rotationally invariant temporal patterns (TPs), and the Synaptic Kernel Adaptation Network (SKAN), a neural network capable of unsupervised learning of a spatio-temporal pattern of input spikes. Our results demonstrate rapid learning and recognition of simple hand gestures with no prior training and minimal usage of FPGA hardware.