Research map / Video / Action recognition
Anonymization for Skeleton Action Recognition
Saemi Moon, Myeonghyeon Kim, Zhenyue Qin, Yang Liu, Dongwoo Kim
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 |
|---|---|---|---|
| Anonymize skeletons using the adversarial framework with a residual anonymizer | Shift-GCN on NTU60 · NTU60 | Re-identification accuracy: 0.8095 -> 0.0420 | Improved |
| Anonymize skeletons using the adversarial framework with a residual anonymizer | Shift-GCN on NTU60 · NTU60 | Action accuracy: 0.9510 -> 0.9175 | Hurt |
| Use an adversarial attack rather than the general anonymization framework | Shift-GCN on NTU60 · NTU60 | Re-identification accuracy on a non-attacked model: 0.0000 -> 0.3621 | Mixed |
| Add random noise with increasing variance | Shift-GCN on NTU60 · NTU60 | Action accuracy: 0.7565 -> 0.0450 | Hurt |
| Increase β from 5 to 75, strengthening reconstruction regularization (ablation) | Shift-GCN with residual anonymizer · NTU60 | Re-identification accuracy: 0.08016 -> 0.15790 | Hurt |
| Increase β from 5 to 75, strengthening reconstruction regularization (ablation) | Shift-GCN with residual anonymizer · NTU60 | Action accuracy: 0.9144 -> 0.9385 | Improved |
| Use a larger reconstruction error to reduce re-identification accuracy (ablation) | Residual anonymizer on NTU60 · NTU60 | Action accuracy | Mixed |
| Use U-Net instead of the residual anonymizer (ablation) | Shift-GCN identity anonymization · NTU60 | Re-identification accuracy: 0.04202 -> 0.05701 | Hurt |
| Use U-Net instead of the residual anonymizer (ablation) | Shift-GCN gender anonymization · ETRI-activity3D | Gender accuracy: 0.5875 -> 0.5671 | Improved |
| Increase α while holding β fixed (ablation) | Shift-GCN with residual anonymizer on NTU60 · NTU60 | Action accuracy | Hurt |
| Increase α while holding β fixed (ablation) | Shift-GCN with residual anonymizer on NTU60 · NTU60 | Re-identification accuracy | Mixed |
| Anonymize gender with the proposed method | Shift-GCN on ETRI-activity3D · ETRI-activity3D | Gender accuracy | Improved |
Closest papers by meaning
- Learning to Anonymize Faces for Privacy Preserving Action Detection (ECCV 2018 · similarity 0.87)
- DisentangledTMR: Privacy-Preserving Skeleton Motion Retargeting via Factorized Transformers (ECCV 2026 · similarity 0.87)
- Privacy-Preserving Deep Action Recognition: An Adversarial Learning Framework and A New Dataset (TPAMI 2020 · similarity 0.86)
- Privacy-centric Deep Motion Retargeting for Anonymization of Skeleton-Based Motion Visualization (ICCV 2025 · similarity 0.86)
- Skeleton-based Action Recognition with Non-linear Dependency Modeling and Hilbert-Schmidt Independence Criterion (AAAI 2025 · similarity 0.84)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 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.
