Computer vision is the problem of understanding imagery from pixels. We are surrounded by billions of cameras, collecting high resolution video and imagery from diverse sources such as security, social media, aerial and satellite platforms. These images contain actionable information, however, the volume of media generated by these cameras far exceeds the number of analysts needed for effective monitoring. Computer vision and machine learning research based on deep convolutional networks has achieved, or surpassed, human level accuracy in recognizing objects and faces in unconstrained scenes, and STR is developing technology in collaboration with academic partners to further extend this state of the art. STR is leading research efforts in the full stack of video and image processing tasks from (i) 4D modeling to reconstruct 3D models of dynamic scenes to provide situational awareness in large operational areas, (ii) object detection and tracking to provide localization of relevant objects in scenes, (iii) face recognition in the wild to identify subjects, (iv) activity detection to identify and detect threats to national security and (v) adversarial modeling to better understand how to make convolutional networks robust to attack.