NeuroPsense SIGNED

Project "NeuroPsense" data sheet

The following table provides information about the project.

 Partnership

Take a look of project's partnership.

Map

 Project objective

Neuromorphic computing has demonstrated high potential for creating computing systems with order-of-magnitude improvements in energy efficiency and robustness to noisy or unreliable sensory signals, such as those inherent in vision. However, a significant roadblock to realizing the full potential of this emerging brain-inspired technology is the current practical need to use inefficient and slow (high latency) legacy von Neumann architectures to convert the input data that needs to be processed, and supply it to the neuromorphic system for further processing. A promising solution to this problem is the recent availability of state-of-the-art neuromorphic sensors, which produce asynchronous event-based output in a form for neuromorphic processing. In parallel, we have developed state-of-the-art neuromorphic processors in the ERC NeuroP project, opening the path to creating fully neuromorphic combined sensing and processing systems. Here we will demonstrate the potential of this technology by building a proof of concept Neuromorphic Sensory Processor (NSP), which will directly interface the neuromorphic Dynamic Vision Sensor with one of the neuromorphic processor devices developed in the ERC NeuroP project. This will represent the first ever general-purpose, end-to-end, fully neuromorphic vision sensing and processing system available for general usage. In this project we will build a technology demonstrator and a detailed commercial business case for this technology, and demonstrate both its technological and commercial advantages. Possible applications for the technology include ultra-high performance and ultra-low power visual processing in ambient surveillance, driver assistance, mobile/wearable devices and robotics.