imageomics/GZCD
收藏资源简介:
GZCD(格氏斑马普查数据集)是一个生态计算机视觉数据集,专为野生动物检测、物种分类和个体动物重识别(Re-ID)而设计。该数据集完全在野外采集,重点关注濒危的马科和有蹄类动物种群,特别是格氏斑马、平原斑马和网纹长颈鹿。数据集包含在肯尼亚梅鲁县拍摄的野外照片,由13名摄影师在2016年和2018年的“大格氏斑马普查活动”(GGR)期间历时四天拍摄。数据收集由公民科学家和环保主义者在25,000平方公里的范围内进行。该数据集旨在帮助训练和测试新的计算机视觉算法,最初是Jason Parham博士论文工作的一部分,后由Imageomics研究所团队完善。数据集提供原始生态图像数据以及经过GGR管道精心处理的输出,包括边界框、视角、普查注释分数和个体动物重识别簇的注释。图像通过9阶段GGR管道处理,强调个体动物身份识别,适用于计算机视觉和生态学交叉研究。
The GZCD (Grevys Zebra Census Dataset) is an ecological computer vision dataset designed for wildlife detection, species classification, and individual animal re-identification (Re-ID). Captured entirely in the field, this dataset emphasizes endangered equid and ungulate populations, specifically Grevys zebras, plains zebras, and reticulated giraffes. GZCD consists of field photographs captured in Meru County, Kenya, taken over four days by 13 photographers during the 2016 and 2018 iterations of the Great Grevys Rally (GGR). The data collection was carried out by citizen scientists and conservationists across a 25,000 square kilometer range. It was curated to help train and test new computer vision algorithms, originally as part of Jason Parhams PhD dissertation work and refined by a team at the Imageomics Institute. The dataset provides raw ecological image data alongside the carefully processed outputs of the GGR Pipeline, featuring annotations for bounding boxes, viewpoints, census annotation scores, and individual animal Re-ID clusters processed through the 9-stage GGR pipeline.




