Custom Based Obstacle Detection Using Yolo v3 for Low Flying Drones
N Aswini, Uma S V · 2021
Low flying drones are on great demand now a days due to their vast applications in Security, Surveillance, Construction sites, Emergency Response, logistics, firefighting, Precision agriculture, Traffic monitoring, Crowd monitoring etc. The greatest advantage is that there is no risk of human life! These unmanned vehicles facilitate safe and secure inspections in complicated and dangerous environments. A fully autonomous drone needs support of various sensors such as Lidars, Radars, ultrasonic sensors, Inertial Measurement Units, Global Positioning system and cameras to detect and avoid possible obstacles in the navigation path. In this paper, we are trying to find a possible solution to detect custom based obstacles using a camera as the primary sensing element. Transfer learning is performed using a YOLO v3 deep learning network.