"Airborne Object Detection and Threat Analysis (AODTA) Dataset"

"Abstract\u2014With the growing use of airborne objects such as commercial aircraft, drones, and UAVs, there is an urgent need for real-time automated threat assessment. Existing methods rely heavily on manual observation, leading to inefficiencies. In this work, we propose a dual-task EfficientNetB4-based model for airborne object classification and threat-level prediction. Due to the lack of a clean and well-balanced dataset, we developed the AODTA Dataset by integrating multiple public sources. We compared performance on both the AVD Dataset and our proposed AODTA Dataset, and also evaluated ResNet-50, which under performed compared to EfficientNetB4. The EfficientNetB4 model achieved 96% object classification accuracy and 90% threat-level prediction accuracy, demonstrating strong potential for surveillance, defense, and airspace management applications. While the title refers to \u201cdetection,\u201d in this work, we focus specifically on airborne object classification and threat-level prediction, using already localized object images provided by public datasets."