A Survey of Deep Learning Application in Dynamic Visual SLAM

Using the camera as the main device for collecting environmental information, visual simultaneous localization and mapping (SLAM) aims to determine the robot's current location and build its surroundings. SLAM system works well in static environments, but it doesn't work well enough in dynamic environments. With many advances in computer vision achieved through deep learning approach, and considering SLAM problem is similar with the computer vision tasks, there is a growing tendency among scholars to consider how to solve the problem of SLAM by using deep learning methods, and put forward some good solutions. The goal of this article is to briefly introduce the SLAM system first and then introduce the dynamic SLAM system in detail proposed in recent years.