2026-10-14T16:01:16Z https://www.scidb.cn/oai

10.57760/sciencedb.noda.0000k 2026-09-28T10:08:56Z

2026-09-28 Tibetan Antelope UAV Object Detection Dataset for Representative Regions of the Qinghai-Xizang Plateau doi:10.57760/sciencedb.noda.0000k en This dataset is a high-resolution UAV object detection dataset for Tibetan antelopes in representative alpine ecosystems of the Qinghai-Xizang Plateau. It focuses on automated Tibetan antelope detection and individual counting, while also including sympatric animals such as kiang, yak, and domestic sheep. The data were collected through multiple UAV field surveys conducted in the Qiangtang and Hoh Xil nature reserves and their surrounding areas on the Qinghai-Xizang Plateau. The surveys spanned approximately one and a half years and covered all four seasons—spring, summer, autumn, and winter—as well as a variety of representative plateau environments, including alpine meadows, deserts, bare ground, riverbanks, saline-alkali areas, and seasonally snow-covered regions.The dataset contains 4,946 original large-format, high-resolution UAV images without cropping and more than 140,000 annotated animal instances. The imagery includes both nadir and multi-angle oblique views. Unlike datasets that provide only pre-cropped image patches containing individual targets, this dataset preserves the complete UAV images together with their surrounding contextual information, thereby more faithfully representing the spatial structure, target distribution, and complex background characteristics encountered in real-world UAV wildlife surveys.Owing to variations in animal body size, aggregation behavior, UAV viewing angle, flight altitude, and complex plateau backgrounds, the dataset exhibits pronounced characteristics of small targets, dense distributions, spatial clustering, close inter-object proximity, and partial occlusion. Some images capture large-scale animal aggregations, with more than 2,000 animal instances in a single image. Different animal categories exhibit clear differences in target scale, spatial density, body shape, and appearance, while also showing a certain degree of inter-class visual similarity. The dataset can therefore support research on small-object detection, dense-object detection, discrimination among visually similar animal species, and cross-scene generalization under challenging alpine conditions.All animal targets were manually annotated, reviewed, and subjected to quality control, with instance-level bounding-box annotations provided. The dataset can be used for research and applications including automated Tibetan antelope detection and individual counting, wildlife-livestock discrimination, dense small-object detection, and intelligent UAV-based wildlife patrols. To protect wildlife and ensure the security of sensitive habitats, the publicly released version does not contain precise geographic coordinates. Tibetan Antelope; Qinghai-Xizang Plateau; UAV Imagery; Wildlife Monitoring; Object Detection Wang Guizhou He Guojin Zhou Dengji Yin Ranyu An Shihao Cui Suxiang Zhang Zhaoming He Chuan EMBARGO https://creativecommons.org/licenses/by-nc-sa/4.0/ dataset Science Data Bank