Drones and Unmanned Aerial Vehicles (UAVs) are one of the cheapest and most difficult to counter aerial vehicles inducing security and safety concerns in many sectors. Recent events suggest that traditional detection technologies such as radar are not effective for detecting these vehicles despite the huge investments from countries suffering from various drone attacks. Current technologies often rely on data such as images to detect and locate drones in the sky using advanced AI-based approaches. Given these challenges, the global aim of this research is to evaluate the effectiveness of various deep learning-based object detection methods on the task of visually detecting airborne drones from camera feeds. We mostly compare single-stage, two-stage, and anchor-free object recognition methods in terms of recognition performance and speed. As per the data, we use as a common benchmark; the Anti-UAV dataset presented at the ICCV 2021 conference. From the results, we see that anchor-free models perform better compared to the other methods.
Evaluation of Single-Stage, Two-Stage, and Anchor-Free Object Recognition Methods for UAV Detection
Lect. Notes in Networks, Syst.
01.01.2023
9 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Analysis of Anchor-Based and Anchor-Free Object Detection Methods Based on Deep Learning
British Library Conference Proceedings | 2020
|