Aerial UAV Thermal – Inferred Unified is a multi-source object detection dataset that integrates aerial drone imagery with thermal infrared imaging, designed to address the fragmentation of thermal and aerial data sources and inconsistent annotation formats. It is suitable for tasks such as drone object detection, thermal infrared imaging, small object detection, personnel and vehicle recognition, drone detection, and mine or weapon detection.

The dataset contains 27,925 images and 90,448 object instances, consolidating 6 independent source datasets and unifying them into 7 categories: Person, Car, Bicycle, Other Vehicle, Drone/UAV, Mine, and Gun. It is split into training, validation, and test sets, retaining the original source information for each sample, and supports independent separation and analysis by source dataset.

Citation

@misc{umut_uygur_2026,
title={Aerial UAV Thermal – Inferred Unified Dataset},
url={https://www.kaggle.com/dsv/18977569},
DOI={10.34740/KAGGLE/DSV/18977569},
publisher={Kaggle},
author={Umut Uygur},
year={2026}
}