This proposal addresses the use of the HBP spiking hardware platforms to explore and exploit neuromorphic spiking computational architectures with feedback for high speed vision. We will combine the new spiking Dynamic Vision Sensors (DVS) with novel event-driven computing paradigms. The aim is to use a pair of DVS sensors for stereo based object recognition, with attentional feedback for recognizing and tracking objects at very high speeds. Spike-based (also called 'event-driven' (ED)) sensing and processing allows to perform multi-Iayer signal processing (with feedback), while the sensor is observing reality. This contrasts with conventional computer vision where a sensor first acquires frames and afterwards a computing system processes the frames.