MUVY: A multi-view user-generated sports video dataset
Larissa Luzia Soares Pessoa, Elton Alencar, Rosiane de Freitas · Data in Brief · 2026
The MUVY dataset contains user-generated video recordings of real-world sporting events captured from multiple viewpoints using handheld mobile devices within sports venues. The dataset was constructed from publicly available Creative Commons-licensed YouTube videos and organized to preserve the multi-view relationships between independently recorded camera perspectives of the same sporting action. MUVY comprises 141 video recordings grouped into 31 events across seven sports categories, including athletics, soccer, basketball, American football, artistic gymnastics, tennis, and cricket, totaling approximately 3 h and 15 min of footage. Each recording is accompanied by video-level metadata and frame-level spatial annotations, including manually assigned camera viewpoint labels and object detection annotations identifying players, referees, goalkeepers, and sports ball classes. The dataset is organized in a hierarchical structure by sport category, event identifier, and camera recording, providing access to the original videos, extracted image frames, and associated annotation files. MUVY may support research on multi-view video analysis tasks that involve spatial reasoning and viewpoint-aware processing in unconstrained acquisition conditions.