Research map / Video / Action recognition
Skeleton Cloud Colorization for Unsupervised 3D Action Representation Learning
Siyuan Yang, Jun Liu, Shijian Lu, Meng Hwa Er, Alex C. Kot
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 |
|---|---|---|---|
| add colorization pretraining (temporal stream) vs random encoder (ablation) | linear classifier · NTU RGB+D cross-subject unsupervised | accuracy: 61.8 -> 71.6 | Improved |
| add spatial colorization stream to the temporal stream (ablation) | TS colorization · NTU RGB+D cross-subject unsupervised | accuracy: 71.6 -> 74.6 | Improved |
| add person-level stream to TS+SS (ablation) | TS+SS colorization · NTU RGB+D cross-subject unsupervised | accuracy: 74.6 -> 75.2 | Improved |
| add person-level stream to TS+SS (ablation) | TS+SS colorization · NTU RGB+D cross-view unsupervised | accuracy: 82.6 -> 83.1 | Improved |
| add colorization pretraining vs random init (ablation) | encoder · NW-UCLA unsupervised | accuracy: 78.6 -> 90.1 | Improved |
| add colorization pretraining vs random init (ablation) | encoder · NTU RGB+D cross-subject supervised | accuracy: 76.5 -> 84.2 | Improved |
| add colorization pretraining vs random init (ablation) | encoder · NTU RGB+D cross-subject 1% labels | accuracy: 27.1 -> 42.9 | Improved |
| add colorization pretraining vs random init (ablation) | encoder · NW-UCLA 10% labels | accuracy: 54.9 -> 71.3 | Improved |
| add person-level stream to TS+SS (ablation) | TS+SS colorization · NTU RGB+D cross-subject supervised | accuracy: 86.3 -> 88.0 | Improved |
| add person-level stream alone as the only colorization (ablation) | PS colorization · NTU RGB+D cross-subject unsupervised | accuracy: 61.8 -> 64.2 | Improved |
| add person-level stream to TS+SS in semi-supervised training (ablation) | TS+SS colorization · NTU RGB+D cross-subject 5% labels | accuracy: 64.7 -> 65.7 | Improved |
| colorize 50% of points (odd frames) instead of other ratio (ablation) | TS colorization · NTU RGB+D cross-subject unsupervised | accuracy: 65.7 -> 71.6 | Improved |
Closest papers by meaning
- Self-Supervised 3D Action Representation Learning With Skeleton Cloud Colorization (TPAMI 2023 · similarity 0.97)
- Unsupervised Representation Learning With Long-Term Dynamics for Skeleton Based Action Recognition (AAAI 2018 · similarity 0.90)
- PREDICT & CLUSTER: Unsupervised Skeleton Based Action Recognition (CVPR 2020 · similarity 0.89)
- SkeletonMAE: Graph-based Masked Autoencoder for Skeleton Sequence Pre-training (ICCV 2023 · similarity 0.89)
- Heterogeneous Skeleton-Based Action Representation Learning (CVPR 2025 · similarity 0.88)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (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.
