Building a cloud-based ‘visual engine’ that allows camera devices to understand their environment, using computer vision.