Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Prompt-Guided Zero-Shot Anomaly Action Recognition Using Pretrained Deep Skeleton Features
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| add the prompt-guided action score to the OoD score (ablation) | Ours · RWF-2000 | accuracy (%): 71.8 -> 82.5 | Helped |
| add the OoD score to the prompt-guided action score (ablation) | Ours · RWF-2000 | accuracy (%): 80.0 -> 82.5 | Helped |
| add the OoD score to the prompt-guided action score (ablation) | Ours · Kinetics-250 random split | ROC-AUC: 0.52 -> 0.69 | Helped |
| add the OoD score to the prompt-guided action score (ablation) | Ours · Kinetics-250 meaningful split | ROC-AUC: 0.60 -> 0.79 | Helped |
| replace the proposed architecture with StructPool (ablation) | Ours · Kinetics-250 meaningful split | ROC-AUC: 0.79 -> 0.78 | Hurt |
| use StructPool with HRNet skeletons (ablation) | Ours · RWF-2000 | accuracy (%): 82.5 -> 90.3 | Helped |
| increase pose detection error ratio (ablation) | Ours and ST-GCN · RWF-2000 | accuracy (%): Ours: 82.5 -> 78.8; ST-GCN: 83.3 -> 52.0 | Mixed |
| use the full anomaly score instead of only the prompt score (ablation) | Ours with prompt “punch, kick or push related to fighting” · RWF-2000 | accuracy (%): 80.0 -> 82.5 | Helped |
| vary the hand-crafted text prompt (ablation) | Ours, full anomaly score · RWF-2000 | accuracy (%): 79.3; 80.8; 81.5; 82.5 | Mixed |
| evaluate across five scene subsets rather than a single scene (ablation) | Ours versus ST-GCN · RWF-2000 | accuracy variance: 0.81 -> 0.08 | Helped |
Closest papers by meaning
- Neuron: Learning Context-Aware Evolving Representations for Zero-Shot Skeleton Action Recognition (CVPR 2025 · similarity 0.87)
- Semantic-guided Cross-Modal Prompt Learning for Skeleton-based Zero-shot Action Recognition (CVPR 2025 · similarity 0.86)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 2025 · similarity 0.86)
- Recovering Complete Actions for Cross-dataset Skeleton Action Recognition (NeurIPS 2024 · similarity 0.86)
- Heterogeneous Skeleton-Based Action Representation Learning (CVPR 2025 · similarity 0.85)
- Stitch, Contrast, and Segment: Learning a Human Action Segmentation Model Using Trimmed Skeleton Videos (AAAI 2025 · similarity 0.85)
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.