Research map / Video / Action recognition
Skeleton Motion Words for Unsupervised Skeleton-Based Temporal Action Segmentation
Uzay Gökay, Federico Spurio, Dominik R. Bach, Juergen Gall
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 |
|---|---|---|---|
| keep joints disentangled in the embedding instead of concatenating (ablation) | SMQ · LARa | MoF: 26.6 -> 37.4 | Improved |
| group latents into one-second patches instead of frame-wise quantization (ablation) | SMQ · LARa | MoF: 33.9 -> 37.4 | Improved |
| add commitment loss (ablation) | SMQ · LARa | MoF: 31.0 -> 37.4 | Improved |
| add reconstruction loss (ablation) | SMQ · LARa | MoF: 29.9 -> 37.4 | Improved |
| use inter-joint distance MSE instead of plain MSE (ablation) | SMQ · LARa | MoF: 34.6 -> 37.4 | Improved |
| use inter-joint distance MSE instead of root distance MSE (ablation) | SMQ · LARa | MoF: 28.0 -> 37.4 | Improved |
| make the decoder joint-independent as well as the encoder (ablation) | SMQ · LARa | MoF: 35.9 -> 37.4 | Improved |
| make the encoder joint-independent with a joint-independent decoder (ablation) | SMQ · LARa | MoF: 28.1 -> 37.4 | Improved |
| use MS-TCN autoencoder instead of ST-GCN (ablation) | SMQ · LARa | MoF: 34.2 -> 37.4 | Improved |
| add orientation angles to position input (ablation) | SMQ · LARa | MoF: 33.9 -> 37.4 | Improved |
| initialize codebook with time series k-means instead of random (ablation) | SMQ · LARa | MoF: 37.4 -> 37.3 | No clear effect |
| patch size 75 instead of 50 (ablation) | SMQ · LARa | MoF: 37.4 -> 37.2 | No clear effect |
Closest papers by meaning
- Stitch, Contrast, and Segment: Learning a Human Action Segmentation Model Using Trimmed Skeleton Videos (AAAI 2025 · similarity 0.92)
- Heterogeneous Skeleton-Based Action Representation Learning (CVPR 2025 · similarity 0.89)
- Recovering Complete Actions for Cross-dataset Skeleton Action Recognition (NeurIPS 2024 · similarity 0.89)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 2025 · similarity 0.88)
- Neuron: Learning Context-Aware Evolving Representations for Zero-Shot Skeleton Action Recognition (CVPR 2025 · similarity 0.88)
- 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.
