Research map / Video / Action recognition
SAFER-Activities: A Dataset for Smart Assessment of Fall Events and Routine Activities
Diwas Lamsal, Pramod Wickramatilake, Jednipat Moonrinta, Mongkol Ekpanyapong, Matthew Dailey
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 2D skeleton instead of frozen RGB features | MS-G3D / VideoMAE · SAFER-Activities In-Lab | macro F1: VideoMAE 80.3 vs MS-G3D 87.2 | Improved |
| evaluate frozen RGB under domain shift (In-Lab to Non-Lab) | DINOv3 / CLIP / VideoMAE · SAFER-Activities | macro F1: DINOv3 72.1 -> 23.8; CLIP 70.6 -> 35.6; VideoMAE 80.3 -> 48.7 | Hurt |
| evaluate skeleton model under domain shift (In-Lab to Non-Lab) | PoseC3D · SAFER-Activities | macro F1: 86.6 -> 79.3 | Hurt |
| fuse frozen RGB with CNN1D skeleton by concatenation | CNN1D · SAFER-Activities Wheelchair | macro F1: 68.2 -> 78.6 | Improved |
| fuse frozen RGB with CNN1D skeleton by concatenation | CNN1D · SAFER-Activities Non-Lab OOD | macro F1: 73.8 -> 32.9 (DINOv3), 54.5 (VideoMAE) | Hurt |
| fuse frozen RGB with CNN1D skeleton by concatenation | CNN1D · SAFER-Activities In-Lab | macro F1: 83.8 -> 85.8 (VideoMAE) | Improved |
| replace concatenation with modality dropout (ModDrop) (ablation) | fusion · SAFER-Activities Non-Lab OOD | macro F1: VideoMAE 54.5 -> 67.3; DINOv3 32.9 -> 59.6 | Improved |
| replace concatenation with QMF or OGM-GE (ablation) | fusion · SAFER-Activities Non-Lab OOD | macro F1: DINOv3 32.9 -> 32.5 (QMF), 29.0 (OGM-GE) | No clear effect |
| use MMCL (skeleton only at inference) (ablation) | CNN1D fusion · SAFER-Activities Wheelchair | macro F1: 68.2 -> 69.7 (CLIP) | Improved |
| increase temporal stride to 3 (144-frame context) (ablation) | 2D skeleton models · SAFER-Activities Wheelchair | macro F1: ST-GCN++ 71.9 -> 75.6; PoseC3D 75.8 -> 78.3 | Improved |
| increase temporal stride to 3 (ablation) | 2D skeleton models · SAFER-Activities In-Lab | macro F1: MS-G3D 87.2 -> 87.1 | No clear effect |
| test skeleton vs frozen RGB on ImViA fall detection | PoseC3D / CLIP / DINOv3 · ImViA | fall F1: CLIP 0.0 vs PoseC3D 96.8 | Improved |
Closest papers by meaning
- NTU RGB+D 120: A Large-Scale Benchmark for 3D Human Activity Understanding (TPAMI 2019 · similarity 0.85)
- Action Recognition Using Rate-Invariant Analysis of Skeletal Shape Trajectories (TPAMI 2016 · similarity 0.85)
- NTU RGB+D: A Large Scale Dataset for 3D Human Activity Analysis (CVPR 2016 · similarity 0.84)
- HARDVS: Revisiting Human Activity Recognition with Dynamic Vision Sensors (AAAI 2024 · similarity 0.83)
- Inertia-1: An Open Exploration of Wearable Motion Foundation Models (NeurIPS 2026 · similarity 0.83)
- AI-Driven Personalized Fall Prevention for Older Adults (AAAI 2025 · 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.
