Research map / Video / Action recognition
Self-Supervised Action Representation Learning from Partial Spatio-Temporal Skeleton Sequences
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 PSTL spatial and temporal masking framework rather than SkeletonBT | SkeletonBT · NTU-60 xsub, linear evaluation | accuracy: 68.5 -> 77.3 | Helped |
| add the PSTL spatial and temporal masking framework rather than SkeletonBT | SkeletonBT · NTU-60 xview, linear evaluation | accuracy: 71.6 -> 81.8 | Helped |
| add PSTL masking rather than SkeletonBT | SkeletonBT · PKU-MMD Part II, linear evaluation | accuracy: 38.2 -> 49.3 | Helped |
| add Central Spatial Masking (ablation) | SkeletonBT · NTU-60 xsub, linear evaluation | accuracy: 68.5 -> 71.8 | Helped |
| add Central Spatial Masking (ablation) | SkeletonBT · NTU-60 xview, linear evaluation | accuracy: 71.6 -> 75.4 | Helped |
| add Motion Attention Temporal Masking (ablation) | SkeletonBT · NTU-60 xsub, linear evaluation | accuracy: 68.5 -> 73.7 | Helped |
| add Motion Attention Temporal Masking (ablation) | SkeletonBT · NTU-60 xview, linear evaluation | accuracy: 71.6 -> 78.0 | Helped |
| replace random spatial masking with centrality-weighted CSM while retaining MATM (ablation) | PSTL masking framework · NTU-60 xsub, linear evaluation | accuracy: 75.6 -> 77.3 | Helped |
| replace random temporal masking with motion-attention MATM while retaining CSM (ablation) | PSTL masking framework · NTU-60 xview, linear evaluation | accuracy: 80.4 -> 81.8 | Helped |
| vary the number of CSM-masked joints from the full method's 9 to 5 (ablation) | PSTL, joint stream · NTU-60 xsub | accuracy: 77.3 -> 76.7 | Hurt |
| vary the number of CSM-masked joints from the full method's 9 to 7 (ablation) | PSTL, joint stream · NTU-60 xsub | accuracy: 77.3 -> 76.0 | Hurt |
| vary MATM top-K from the full method's 10 to 6 (ablation) | PSTL, joint stream · NTU-60 xsub | accuracy: 77.3 -> 75.3 | Hurt |
Closest papers by meaning
- Exploring Adaptive Masked Reconstruction for Self-Supervised Skeleton-Based Action Recognition (CVPR 2026 · similarity 0.91)
- MaskCLR: Attention-Guided Contrastive Learning for Robust Action Representation Learning (CVPR 2024 · similarity 0.89)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 2025 · similarity 0.88)
- SCD-Net: Spatiotemporal Clues Disentanglement Network for Self-Supervised Skeleton-Based Action Recognition (AAAI 2024 · similarity 0.88)
- Stitch, Contrast, and Segment: Learning a Human Action Segmentation Model Using Trimmed Skeleton Videos (AAAI 2025 · similarity 0.88)
- Heterogeneous Skeleton-Based Action Representation Learning (CVPR 2025 · similarity 0.87)
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.