Research map / Video / Action recognition
Modeling the Relative Visual Tempo for Self-supervised Skeleton-based Action Recognition
Yisheng Zhu, Hu Han, Zhengtao Yu, Guangcan Liu
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 |
|---|---|---|---|
| replace tempo prediction with relative visual tempo learning (ablation) | RVTCLR · NTU-60 Xsub linear joint | top-1 accuracy: 52.7 -> 63.1 | Improved |
| add appearance-consistency task to RVTL (ablation) | RVTCLR · NTU-60 Xsub linear joint | top-1 accuracy: 63.1 -> 73.3 | Improved |
| add nearest neighbor mining stage (ablation) | AC+RVTL · NTU-60 Xsub linear joint | top-1 accuracy: 73.3 -> 74.4 | Improved |
| add Distribution-Consistency branch on motion stream (ablation) | RVTCLR · NTU-60 Xsub linear motion | top-1 accuracy: 57.5 -> 70.8 | Improved |
| add DC branch on joint stream (ablation) | RVTCLR · NTU-60 Xview linear joint | top-1 accuracy: 79.4 -> 79.1 | No clear effect |
| use Gaussian noise, blur and channel mask as strong augmentations instead of normal (ablation) | SDA in DC branch · NTU-60 Xsub linear motion | top-1 accuracy: 55.2 -> 67.9 | Improved |
| add spatial flip or temporal flip or rotate as strong augmentation (ablation) | SDA · NTU-60 Xview linear motion | top-1 accuracy: 58.2 -> 55.9 (temporal flip) | Hurt |
| add IIPB and NLB (FSEM) (ablation) | DC branch · NTU-60 Xsub linear motion | top-1 accuracy: 67.9 -> 70.8 | Improved |
| use DD loss with weak-view distribution instead of one-hot targets (ablation) | DC branch · NTU-60 Xsub linear motion | top-1 accuracy: 64.6 -> 70.8 | Improved |
| RVTCLR+ vs 3s-AimCLR under linear eval on NTU-120 Xsub | 3s-RVTCLR+ · NTU-120 Xsub linear | top-1 accuracy: 68.2 -> 68.0 | No clear effect |
Closest papers by meaning
- Self-Supervised Skeleton-Based Action Representation Learning: A Benchmark and Beyond (IJCV 2026 · similarity 0.90)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 2025 · similarity 0.90)
- LAC - Latent Action Composition for Skeleton-based Action Segmentation (ICCV 2023 · similarity 0.90)
- Stitch, Contrast, and Segment: Learning a Human Action Segmentation Model Using Trimmed Skeleton Videos (AAAI 2025 · similarity 0.89)
- Neuron: Learning Context-Aware Evolving Representations for Zero-Shot Skeleton Action Recognition (CVPR 2025 · similarity 0.89)
- Part-aware Unified Representation of Language and Skeleton for Zero-shot Action Recognition (CVPR 2024 · similarity 0.88)
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.
