Colloquium: Federico Paredes Vallés (C&O)
19 juni 2018 14:00 - Locatie: Meeting room 1 (NB 2.51), Faculty of Aerospace Engineering, Kluyverweg 1, Delft.
Neuromorphic Computing of Event-based Data for Vision-based Navigation
As part of a computational framework with the potential of resembling the main functionalities of biological brains, Spiking Neural Networks offer an efficient alternative to artificial solutions in the task of optical flow estimation. This paper presents the first hierarchical spiking architecture in which motion selectivity emerges in a biologically-plausible unsupervised fashion from the stimuli generated with an event-based camera. A novel adaptive neuron model and Spike-Timing-Dependent Plasticity formulation are at the core of this network governing its spike-based processing and learning, respectively. After convergence, the neural architecture exhibits the main properties of biological visual motion systems: feature extraction and local and global motion perception. In addition, a readout mechanism is introduced for the estimation of optical flow visual observables from the activity of the network. Experimental results show that, despite the spiking nature of the processing, accurate ventral flow estimates can be obtained over a wide range of speeds.