Frame level labels support small UAV point detection but have limits
What Frame-Level Labels Can and Cannot Do for Small-UAV Point Detection in Thermal Video
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
Detecting small drones in thermal video is hard, especially when detailed location labels are not available for training detection models. The authors studied a method that learns to find drone locations using only labels that say if a drone is present in a video frame or not. They found that this approach can produce useful location guesses but struggles when the drone signal is weak or when tested on new datasets. Their work helps show when simple labels are enough and when more detailed information is needed for reliable drone detection.
What this means in practice
- •For security operations teams: Use frame-level labels to train small UAV detectors in thermal video when bounding-box annotations are unavailable or costly, improving detection capabilities with limited labeling.
- •For airborne object tracking teams: Employ frame-level label training for point detection of small UAVs in visible-light video to support airborne object tracking where spatial labels are scarce.
Authors
Wonbin Son, Gyumum Choi, Junil Seo, Hyungjoon Kim
Abstract
The growing use of unmanned aerial vehicles (UAVs) has increased the importance of image-based UAV detection. Learning-based detectors are trained on imagery and annotations, with annotation type determining the information available during training. We focus on learning localization from frame-level target presence/absence labels when sensor or scene changes make spatial annotations for additional training burdensome. We analyze the detection capability, learning behavior, and potential applications of an existing architecture for point detection of small UAVs, trained with presence/absence labels and requiring no external detector. The architecture freezes spatial features learned through classification and trains a readout with the same frame labels to produce spatial score maps and point detections. On two thermal infrared datasets, CST Anti-UAV and Anti-UAV410, we evaluate localization hit rates and detection rates under false-alarm constraints, analyze the effects of training stages, label allocation, synthesis, and model configuration, and compare with bounding-box detectors. We also explore potential applications on Airborne Object Tracking (AOT) using its visible-light imagery and frame labels. Classification training strengthened target-related spatial responses, while readout training helped extract them consistently. Distributing similar label counts across more videos yielded higher localization hit rates, while synthesis effects varied by dataset and evaluation criterion. Higher localization hit rates did not always improve detection under false-alarm constraints, and failures remained when target signals were weak relative to background variation and under cross-dataset transfer. These findings provide guidance on label allocation, spatial representations and readouts, synthesis, and false-alarm control.