Towards Airborne Object Detection: A Deep Learning Analysis

TL;DR

This work addresses real-time airborne threat assessment by proposing a dual-task CNN based on EfficientNetB4 that simultaneously classifies airborne objects and predicts threat levels on pre-localized crops. It introduces the AODTA dataset, crafted by aggregating diverse CC0 sources, and benchmarks performance against the AVD dataset and a ResNet-50 baseline. The model achieves 96% class accuracy and 90% threat accuracy on AODTA, outperforming ResNet-50 and validating the feasibility of combined classification and threat inference for surveillance and airspace management. Although not an end-to-end detector, the framework can integrate with detectors like YOLO or Faster R-CNN, providing a practical step toward real-time airborne threat assessment and defense applications.

Abstract

The rapid proliferation of airborne platforms, including commercial aircraft, drones, and UAVs, has intensified the need for real-time, automated threat assessment systems. Current approaches depend heavily on manual monitoring, resulting in limited scalability and operational inefficiencies. This work introduces a dual-task model based on EfficientNetB4 capable of performing airborne object classification and threat-level prediction simultaneously. To address the scarcity of clean, balanced training data, we constructed the AODTA Dataset by aggregating and refining multiple public sources. We benchmarked our approach on both the AVD Dataset and the newly developed AODTA Dataset and further compared performance against a ResNet-50 baseline, which consistently underperformed EfficientNetB4. Our EfficientNetB4 model achieved 96% accuracy in object classification and 90% accuracy in threat-level prediction, underscoring its promise for applications in surveillance, defense, and airspace management. Although the title references detection, this study focuses specifically on classification and threat-level inference using pre-localized airborne object images provided by existing datasets.

Figures (13)

• Figure 1: No Threat Aerial Object Image from the dataset (a) Airplane Civilian Aircraft, (b) Drone Civilian Model, (c) Helicopter Civilian Aircraft, (d) UAV Civilian Model vs. Threat Aerial Object Images from the Airplane Military Aircraft, (f) Drone Military Model, (g) Helicopter Military Aircraft, (h) UAV Military Model
• Figure 2: Airborne Object Class Detection Model
• Figure 3: Airborne Object Threat Detection Model
• Figure 4: EfficientNetB4 Model Architecture
• Figure 5: Training vs. Validation Accuracy and Training vs. Validation loss on AVD dataset. Epochs used: 21