2026-10-15T01:36:40Z https://www.scidb.cn/oai

10.57760/sciencedb.011v4 2026-09-23T19:49:09Z

2026-09-23 A Multi-temporal Remote-Sensing Image Dataset for Great Wall Beacon Tower Detection (2010–2025) doi:10.57760/sciencedb.011v4 en This dataset was developed for the remote-sensing object detection of Great Wall beacon towers. It covers Great Wall beacon towers and their surrounding buffer zones in Shanxi, Hebei, and Shaanxi Provinces, China. The imagery covers multiple discrete years between 2010 and 2025. The images were obtained from Google Earth at zoom level 18, with an approximate ground sampling distance of 0.45 m.The dataset consists entirely of positive samples containing Great Wall beacon-tower targets. It contains 28,367 image–annotation pairs with a total size of approximately 21.90 GB. Specifically, the Shanxi subset contains 9,420 image–annotation pairs and 15,807 bounding boxes; the Hebei subset contains 13,875 pairs and 16,445 bounding boxes; and the Shaanxi subset contains 5,072 pairs and 6,257 bounding boxes. In total, 38,509 valid rectangular bounding boxes are included. No negative samples are included in this dataset.The images are stored in TIFF format, and the annotations are provided in LabelMe JSON format. Each image is paired with a JSON annotation file sharing the same base filename. Every JSON file contains at least one valid rectangle annotation with finite coordinate values and non-zero width and height. The files are organized hierarchically by province, year, and original sample group, while preserving the correspondence between images and annotations.This dataset can support the remote-sensing recognition of Great Wall beacon towers, the training and evaluation of object detection models, and small-object detection research in complex land-surface backgrounds. Users should note that the dataset contains positive samples only; suitable negative samples may be added according to the requirements of specific training or real-world evaluation tasks. Great Wall beacon tower; multi-temporal remote-sensing image; bounding-box annotation; LabelMe; object detection; cultural heritage conservation Zhou Huan Hu Qingwu PUBLIC https://creativecommons.org/licenses/by-nc/4.0/ dataset Science Data Bank