Detecting and tracking objects is a popular technology used on drones. Our Deep Vision project identifies and then tracks certain targets, such as dogs and people, in real time by analyzing the video recorded by the drone’s camera. Conventional tracking algorithms can be very limited in accuracy and capability. With the recent development of AI and deep learning technology, team Deep Vision aims to implement an algorithm on an embedded system that features neural networks-based target tracking for drones. The algorithm first trains the system from an existing dataset, and then will be able to recognize different objects in the same category even if the object itself is not in the original dataset. The hardware we use is an embedded system with a dedicated GPU – Nvidia Jetson TX2. We are participating in 2018 DAC System Design Contest.