Research map / Large language models / AI and society
SALSA: A Novel Dataset for Multimodal Group Behavior Analysis
Xavier Alameda-Pineda, Jacopo Staiano, Ramanathan Subramanian, Ligia Batrinca, Elisa Ricci, Bruno Lepri, Oswald Lanz, Nicu Sebe
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use more training data (1% to 10%) for head pose | joint head-body pose classifier · SALSA view 1 head pose | mean angular error (degrees): 45.7 -> 42.2 | Improved |
| use more training data (1% to 10%) for body pose | joint head-body pose classifier · SALSA view 4 body pose | mean angular error (degrees): 54.6 -> 51.7 | Improved |
| estimate head instead of body pose | joint head-body pose classifier · SALSA pose estimation | mean angular error (degrees): 49.3 -> 45.7 | Improved |
| use body orientation instead of head orientation | graph-cut F-formation detection · SALSA F-formation detection (GT) | F1: 0.82 -> 0.83 | Improved |
| use body orientation instead of head orientation | HVFF ent · SALSA F-formation detection (GT) | F1: 0.69 -> 0.73 | Improved |
| switch from Hough voting (HVFF lin) to graph cut | F-formation detection · SALSA F-formation detection (GT head) | F1: 0.63 -> 0.82 | Improved |
| concatenate MFCCs of all badges (frame-based fusion) instead of per-badge features | speaker recognition · SALSA speaker recognition | accuracy | Improved |
| add infra-red data to visual pose estimates (ablation) | graph-cut F-formation detection · SALSA F-formation detection (estimated pose) | F1 | Improved |
| add audio data to visual pose estimates (ablation) | graph-cut F-formation detection · SALSA F-formation detection (estimated pose) | F1 | No clear effect |
| use estimated visual pose instead of ground-truth pose | graph-cut F-formation detection · SALSA F-formation detection | F1 | Hurt |
| move from Poster session to Party session | HJS-PF tracker · SALSA tracking | failure rate (%): 2.6 -> 9.6 | Hurt |
| move from Poster session to Party session | HJS-PF tracker · SALSA tracking | precision (cm): 15.2 -> 20.1 | Hurt |
Closest papers by meaning
- Part-Aware Bottom-Up Group Reasoning for Fine-Grained Social Interaction Detection (NeurIPS 2025 · similarity 0.86)
- Modeling Multimodal Social Interactions: New Challenges and Baselines with Densely Aligned Representations (CVPR 2024 · similarity 0.86)
- JRDB-Social: A Multifaceted Robotic Dataset for Understanding of Context and Dynamics of Human Interactions Within Social Groups (CVPR 2024 · similarity 0.85)
- The Audio-Visual Conversational Graph: From an Egocentric-Exocentric Perspective (CVPR 2024 · similarity 0.84)
- SocialGesture: Delving into Multi-person Gesture Understanding (CVPR 2025 · similarity 0.83)
- MTGS: A Novel Framework for Multi-Person Temporal Gaze Following and Social Gaze Prediction (NeurIPS 2024 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
