Multi-View Tennis Ball Dataset for Trajectory Estimation: Drone and Court Cameras with Annotated Ground Truth

This dataset provides a dual-view tennis stroke video collection designed for the study of ball detection and trajectory estimation. The videos were recorded simultaneously from two perspectives: an aerial drone view and a court-level camera view, offering complementary visual information. Ground truth coordinates of the tennis ball are included, along with annotation files containing predictions generated using YOLO-based object detection, classical image segmentation, and a hybrid approach that combines both methods. These resources enable direct comparison between traditional computer vision and deep learning methods for sports ball tracking.The dataset is distributed in ZIP format and includes MP4 video files and XLSX annotation files. The annotations provide frame-level ground truth positions, detection results, and performance analysis metrics. This dataset can support research in computer vision, machine learning, trajectory prediction, and sports technology applications such as training analytics,  and physics-informed simulations (e.g., Magnus effect modeling).