SALSA

SALSA数据集是一个用于多模态群体行为分析的新型数据集，由特伦托大学信息工程与计算机科学系的研究团队创建。该数据集记录了18名参与者在自然室内环境中超过60分钟的社交互动，包括海报展示和鸡尾酒会等场景。数据集通过四个静态监控摄像头和每个参与者佩戴的社会计量徽章来记录社交互动，徽章包含麦克风、加速度计、蓝牙和红外传感器。除了原始数据外，还提供了关于个人性格以及他们的位置、头部、身体方向和F-形成信息的标注。SALSA数据集旨在通过多模态分析解决社交场景中的遮挡和拥挤问题，精确估计各种行为线索和性格特征，从而推动自动社交互动分析的研究。

The SALSA dataset is a novel dataset for multimodal group behavior analysis, created by a research team from the Department of Information Engineering and Computer Science, University of Trento. It documents over 60 minutes of social interactions among 18 participants in a natural indoor environment, including scenarios such as poster presentations and cocktail parties. The dataset captures social interactions via four static surveillance cameras and sociometric badges worn by each participant; these badges are equipped with microphones, accelerometers, Bluetooth, and infrared sensors. In addition to the raw data, annotations are provided for individual personality traits, as well as their positions, head and body orientations, and F-formation information. The SALSA dataset aims to address the challenges of occlusion and congestion in social scenes through multimodal analysis, accurately estimate various behavioral cues and personality traits, and thereby advance research on automated social interaction analysis.