Foreground Gating and Background Refining Network for Surveillance Object Detection

摘要

Detecting objects in surveillance videos is an important problem due to its wide applications in traffic control and public security. Existing methods tend to face performance degradation because of false positive or misalignment problems. We propose a no