Continuous Tracking and Recognition of Small Objects in Video Streams Based on YOLO and Spatio-Temporal Contextual Memory Networks | Synapse