We propose two attentive models for event-based vision: an algorithm that tracks events activity within the field of view to locate regions of interest and a fully differentiable attention procedure based on DRAW neural model.