LRDDv3: High-Resolution Long-Range Drone Detection Dataset with Range Information and Thermal Data
Knut Peterson, Zaid Mayers, Azmain Yousuf, Priontu Chowdhury, Asher Zaczepinski, Solmaz Arezoomandan, Reihaneh Maarefdoust, David Han · 2026
Read LRDDv3: High-Resolution Long-Range Drone Detection Dataset with Range Information and Thermal Data by Knut Peterson, Zaid Mayers, Azmain Yousuf, Priontu Chowdhury, Asher Zaczepinski, Solmaz Arezoomandan, Reihaneh Maarefdoust, David Han (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-25942.txt.
Introduction
Peterson and seven colleagues release a large dataset for spotting drones at long range. Shared airspace is growing busier, so distant detection matters for safety. The set holds over one hundred thousand high-resolution colour images from 128 flight videos. Range labels cover the whole set, and nearly thirty thousand infrared images pair with colour frames. The authors call it one of the first drone sets to combine very high resolution with paired infrared images.
Catalog record
| CREATOR | Knut Peterson, Zaid Mayers, Azmain Yousuf, Priontu Chowdhury, Asher Zaczepinski, Solmaz Arezoomandan, Reihaneh Maarefdoust, David Han |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 31874 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.25942
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
LRDDv3: High-Resolution Long-Range Drone Detection Dataset with Range Information and Thermal Data Knut Peterson, Zaid Mayers, Azmain Yousuf, Priontu Chowdhury, Asher Zaczepinski, Solmaz Arezoomandan, Reihaneh Maarefdoust, David Han Abstract Unmanned Aerial Vehicles (UAVs) have quickly become common in various airspaces, representing a wide range of applications from recreation flying to commercial photography and package delivery. With the increasing prevalence of UAVs, it becomes critical that both manned and unmanned aircraft can detect UAVs and other flying objects from long range to effectively track movement and ensure safe operation in shared spaces. While several datasets have been introduced for drone detection, the need for expanded high-quality data persists, especially in the area of high-resolution long-range drone data. To address this, we introduce a high-resolution dataset of 102,532 long-range RGB images of drones, sampled at 5 FPS from 128 distinct video clips taken mid flight during 17 different data collection days spread over 8 months to ensure a wide variety of lighting scenarios, flight locations, and background elements. The dataset boasts comprehensive drone range information across the dataset, as well as 29,630 IR images, all paired with RGB counterparts from the base dataset. As one of the first drone detection datasets to leverage 4K image resolution and paired 640x512 IR images, our work represents a significant advancement to enable the detection of drones at long range. For access to the complete dataset, please visit https://research.coe.drexel.edu/ece/imaple/lrddv3/ . I Introduction The use of Unmanned Aerial Vehicles (UAVs) has been exponentially growing for a wide range of applications from recreation flying to commercial photography and package delivery in both urban and rural environments [ 18 , 10 ] . With the use of UAVs becoming more prevalent, it becomes critical that aerial systems can detect UAVs and other flying objects from long range to effectively track movement and avoid collisions. Different manned or unmanned aircraft need to be able to detect UAVs as we are heading toward a future where both must operate together, sometimes in shared spaces. Various advanced detection systems rely on deep learning algorithms, which can be significantly affected by the size and quality of the datasets used in their training. However, many existing datasets showcase drone images that do not reflect the context in which they will be encountered when being operated for commercial or recreational purposes, like varying weather conditions, backgrounds, time of the day, obstacles seen in dense urban environments [ 22 , 21 ] . Fig. 1: Examples of images in the dataset, featuring different challenges such as drones occluded by objects and foliage, drones blending into noisy backgrounds, and drones flying in snowy, rainy, and clear weather conditions. To bridge this gap, we introduce LRDDv3: High-Resolution Long-Range Drone Detection Dataset. Consisting of 102,532 RGB images taken mid flight during 17 different data collection days spread over 8 months, 29,630 IR images paired with RGB counterparts from the base dataset, and comprehensive drone range information, LRDDv3 stands out for its scale, high-resolution, and image diversity. Additionally, the dataset includes a variety of challenges and features to aid in model generalization and enable robust detections, detailed as follows: Thermal Imaging: UAVs have a heat signature that can vary from the elements of their background, such as buildings, trees, cars, sky, etc [ 5 ] . Using thermal imaging, either on its own or in combination with RGB imaging, drone detection can be made more robust through differentiating between the heat signatures of the drone and background elements. Illumination Variation: Change of lighting conditions can significantly change the appearance of drones seen from another drone during a flight [ 19 , 34 ] . This dataset incorporates images taken in varying backgrounds at different times of the day. i.e. dawn, dusk, noon etc. to reflect real world scenarios in drone operations at different times of the day. The variation in illumination level and shadow effects aim to better train algorithms to recognize drones at different times of the day. Diverse Backgrounds: UAVs are expected to operate in a wide range of environments from dense cities to sparsely inhabited small towns or recreational areas. Advanced drone detection systems need to pick up on the varying contexts of the backgrounds and adapt to different background textures, components and complexities [ 19 , 22 , 21 ] . By incorporating images where drones are operating in a dense urban areas or in a more open parks or over a water body, LRDDv3 aims to aid in improving detection accuracy across varying contexts. Weather Conditions: Changes in weather conditions affect UAV visibility and add new contexts for detection [ 23 , 22 , 21 ] . Our work includes images with a wide range of weather conditions from clear skies to grey-cloudy conditions and urban snowy landscapes. Occlusion and Background Blending: While flying in dense urban areas, UAVs often blend into the background or become partially or fully occluded by objects, such as buildings, traffic lights and road signs, cars and other vehicles etc. [ 27 , 21 ] We address this by including images showing scenarios where the drone is partially indistinguishable from its immediate background, testing any algorithm’s capability in drone detection under compromised visibility. Long-Range: As UAV technology has progressed, the speed of UAVs has also increased, making it essential for detection systems to be capable of detecting them at long range [ 11 ] . To aid with this, the dataset includes images of drones at distances from 0-200 meters from the camera, along with range metadata. The resulting instances of drones in the dataset include bounding boxes with areas as small as 12 total pixels, posing a difficult challenge in long-range drone detection. LRDDv3 improves on previous works by combining 4K resolution, IR imaging, drone range annotations, mobile camera views, and diverse lighting, weather, and background conditions into a single, comprehensive dataset. Our work seeks to significantly improve the data quality of drone datasets with diverse images taken in varying scenarios that aim to mimic conditions in which drone detection will be necessary for in-flight operations in the real world, enabling more reliable integration of drones in national and international airspaces. II Related Work - Existing Datasets TABLE I: Overview of current drone detection datasets Dataset # of Images/Videos Data Type Resolution Camera Dynamics Lighting Occlusion Weather Range Multi-Sensor 203k images 90 audio clips RGB, IR, Audio 640x512 (RGB) 320x256 (IR) Static No No Partial Categorical BirDrone 3.3k images RGB 640x640 Various (from internet) Yes No No No VisoDECT 20.9k images RGB, EO-IR 852x480 Static/Handheld Yes No Yes No RWOQ 56.8k images RGB 640x480 Various (from internet) Yes Partial Yes No SynDroneVision 131.3k images RGB 2560x1489 Mobile/Flying Yes No No No DUT Anti-UAV 10k images RGB Variable (240x160 to 5616x3744) Various Yes No Partial No Drone Hunter 58.6k images RGB 1280x720 Handheld Partial No No No FRED 700k images (30 FPS from video) RGB, Event 1280x720 Static Yes No Yes No ARD-MAV 107.5k images (30 FPS from 60 videos) RGB 1920x1080 Mobile/Flying No Yes No No Multi-View Drone Tracking 5 videos, 30.5min RGB Variable (1440x1080 to 4K) Static (Multi-camera) No No Partial No UAV-Detect 20.4k images RGB 640x640 Mobile/Flying Yes No No No Det-Fly 13.3k images RGB 4K Mobile/Flying