Dronescape: A High-Resolution Drone Footage Dataset for Tree Region Segmentation

Dronescape presents a dataset comprising 25 drone videos showcasing vast areas filled with trees, rivers, and mountains. The dataset includes two subsets: 25 videos with tree segmentation and 25 videos without tree segmentation, offering diverse perspectives on the presence and absence of segmented tree regions. The dataset focuses on highlighting the regions containing trees using the SAM (Segment Anything Model) and Track Anything library. Video object tracking and segmentation techniques are utilized to track the regions of trees throughout the dataset. Each video offers diverse perspectives, camera angles, and lighting conditions, providing rich visual content for analysis and research. The dataset includes annotations in the form of masks, indicating the location and extent of tree regions within each frame. This dataset can be utilized for various applications such as object detection, segmentation, tracking algorithms, ecological studies, environmental monitoring, and urban planning projects. Researchers and developers may encounter challenges related to lighting variations, occlusions, camera motion, and accurate tracking of tree regions while working with this dataset

Dronescape数据集包含25段无人机航拍视频，展现了遍布林木、河流与山地的广袤区域。该数据集包含两个子集：25段带有林木分割标注的视频，以及25段无林木分割标注的视频，可为包含分割林木区域与不包含分割林木区域的场景提供多样化的观测视角。本数据集借助Segment Anything Model（SAM）与Track Anything库对视频中的林木区域进行高亮处理，并采用视频目标跟踪与分割技术，对全数据集内的林木区域实现全程跟踪。每段视频均具备多样化的拍摄视角、相机角度与光照条件，可为分析与研究提供丰富的视觉素材。该数据集附带掩码格式的标注信息，用于标注每一帧画面中林木区域的位置与覆盖范围。本数据集可应用于目标检测、图像分割、目标跟踪算法研发、生态学研究、环境监测以及城市规划项目等诸多场景。研究人员与开发者在使用该数据集时，可能会面临光照变化、目标遮挡、相机运动以及林木区域精准跟踪等相关挑战。