Recent technological advances in low power CMOS imaging has lead to a new class of imaging devices such as Cyclops. Cyclops is a low power and complexity imaging sensor that mates with conventional sensor network nodes such as Motes.In this demo we present deep vision. An attribute based system that enables orchestrating visual observation through an attribute based system. In our work, visual computations reside on Cyclops sensors and can be addressed as attributes through declarative queries. We present our network architecture and showcase the potential applications and opportunities as well as outstanding challenges. We call our approach deep vision due to the integration of vision with the finest granularity of the sensor network.