Comparison of Object Detection on Drone Using Tensorflow with Several Architectures

Industrial revolution 4.0 made the world of technology develop more rapidly. Many things that were previously done by humans now become autonomous, one of which is by using drones in various sectors such as agriculture, military, health, and traffic monitoring. Drones that are integrated with cameras make their implementation more extensive. drones that have cameras can be used to detect various kinds of objects. In this research, drones will detect objects in the form of people using the Tensorflow algorithm. Tensorflow has many types of architectures, but this research uses 4 types of architectures namely SSD MobileNet V2 320×320, SSD MobileNet V2 FPNLite 640×640, CenterNet Resnet50 V1 FPN 512×512, and CenterNet Resnet101 V1 FPN 512×512. The architecture that has the best results is CenterNet Resnet50 V1 FPN 512×512 with an accuracy of 94% and an FPS value of 24.39.