Deep learning when applied to video and oriented imagery can be groundbreaking. And great value is realized when location is added to this equation. In this session, we will explore how to use ArcGIS GeoEvent Server with NVIDIA's DeepStream to build an enterprise-wide video analytics workflow, that is able to analyze massive amounts of real-time video surveillance feeds for deep traffic analytics. We will also cover how to build an end-to-end oriented imagery feature extraction pipeline using custom computer vision networks and ArcGIS for a variety of use cases such as road sign extraction, optical character recognition, and road condition classification.