Lightweight UAV Explosion Detection Dataset
AeroBlast-Vision: Lightweight UAV Explosion Detection Dataset Overview This dataset contains image frames and corresponding bounding box annotations for training lightweight deep learning object detection models . It was developed to provide early warning detection of localised anomalies, specifically focusing on combustion and explosions. The dataset is optimized for Unmanned Aerial Vehicles (UAVs) and real-time inference on edge computing devices. Dataset Details Image Count: 8179 Annotation Format: YOLO (.txt files) Classes: 0: Explosion Data Split : 70 % Train, 20% Valid, 10% Test Edge Deployment & Hardware This dataset is specifically st
Resource information
- Data purpose
- Research dataset
- Subject
- Explosion, Drone, Smoke, Lightweight UAV Explosion, Explosion Detection
- Format
- zip
- Original repository
- 21785791
- DOI
- 10.5281/zenodo.21785791
- Credit basis
- Creators in Zenodo record metadata
- Access
- Open files available from the original Zenodo deposit
- Last checked
- Nov 5, 2026
- License
- cc-by-4.0
- Last record update
- Nov 5, 2026
About this resource
AeroBlast-Vision: Lightweight UAV Explosion Detection Dataset Overview This dataset contains image frames and corresponding bounding box annotations for training lightweight deep learning object detection models . It was developed to provide early warning detection of localised anomalies, specifically focusing on combustion and explosions. The dataset is optimized for Unmanned Aerial Vehicles (UAVs) and real-time inference on edge computing devices. Dataset Details Image Count: 8179 Annotation Format: YOLO (.txt files) Classes: 0: Explosion Data Split : 70 % Train, 20% Valid, 10% Test Edge Deployment & Hardware This dataset is specifically structured for edge computing workflows. Models trained on this data have been successfully deployed and evaluated on edge hardware, specifically the NVIDIA Jetson Nano, demonstrating real-time inference capabilities for Command and Control (C2) UAV dashboards. Origin and Methodology Data was sourced by extracting frames from publicly accessible video footage of various combustion and explosive events. The frames were manually cleaned, filtered for quality, and annotated with bounding boxes to train lightweight computer vision models for rapid hazard detection. Author & Affiliation Irum Nigar Wazir The University of Agriculture Peshawar Zaryab Ahmad Khan FMV Isik University Acknowledgments This dataset and the associated research were supported by an official award from the Undergraduate Research Support Program (URSP) provided by the Directorate of Science and Technology (DOST). License This dataset is licensed under the Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0). To view a copy of this license, visit: https://creativecommons.org/licenses/by-nc/4.0/. Copyright Disclaimer This dataset consists of frames extracted from publicly available YouTube videos, curated strictly for academic research and computer vision training. We do not claim ownership of the original video copyrights. If you are a copyright holder and wish for your content to be removed, please contact us at or , and the specific frames will be taken down immediately.