DeepBrain: Cloud-Based Computation Offloading Framework of
Internet-Connected UAVs for Deep Learning Applications
Submitted to: The 2020 annual International conference on Wireless Communications & Mobile Computing (IWCMC 2020)
The Robotics and Internet of Things lab has performed its first tests on 4G connected drones for deep learning applications. In this first demo, we demonstrate a real-time object detection of a 4G connected drone. The drone sends its video stream to a public cloud server on the Internet through a 4G connection. The cloud server dispatches the video frames to a GPU machine, which applies Yolo3 for real-time object detection. In this demo, we have used a pre-trained Yolo network on COCO dataset. The neural network can be tailored to any other use case. The demo was developed by Anis Koubaa, Mahmoud Alahdab, and Anas Kanhouch. We are working on improving this demo in terms of performance and video quality.