Data underlying PhD thesis: AI in the Sky - Advancing Wildlife Survey Methods in Africa with Deep Learning and Aerial Imagery

This dataset supports the PhD research titled <em>“AI in the Sky: Advancing Wildlife Survey Methods in Africa with Deep Learning and Aerial Imagery”</em>. The research aims to improve wildlife monitoring through deep learning and remote sensing. It focuses on object detection and species counting based on aerial surveys over African wildlife reserves. The dataset includes the Aerial Elephant Dataset (AED) annotations, for which bounding boxes in standard VOC format were created to supplement the original point annotations, and an Antelope Dataset provided by African Parks in South Sudan under a research agreement. These annotations support the training and validation of deep learning models such as YOLO, RT-DETR, CenterNet, U-Net, and D2-Net. Supporting scripts for processing, tiling, annotation handling, quality control, and statistical analysis are included to ensure reproducibility.

本数据集支撑题为《云端AI：依托深度学习与航空影像革新非洲野生动物调查方法》的博士研究。该研究旨在通过深度学习与遥感技术优化野生动物监测工作，核心聚焦于非洲野生动物保护区航空调查中的目标检测与物种计数任务。本数据集包含两类标注资源：其一为航空象类数据集（Aerial Elephant Dataset, AED）的标注信息，研究人员为其补充了标准VOC格式的边界框标注，以完善原始的点标注；其二为由南苏丹境内的非洲公园（African Parks）依据研究合作协议提供的羚羊数据集。上述标注可用于支撑YOLO、RT-DETR、CenterNet、U-Net及D2-Net等深度学习模型的训练与验证工作。数据集还附带了用于数据处理、图像分块、标注管理、质量控制及统计分析的配套脚本，以保障研究的可复现性。