Research map / Video / Action recognition
Learning Probabilistic Embeddings for Unsupervised Action Segmentation
Shuai Li, Duc Vu, 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 |
|---|---|---|---|
| add probabilistic embeddings to ASOT | ASOT · Breakfast | MoF: 56.4 -> 60.7 | Improved |
| add probabilistic embeddings to ASOT | ASOT · Desktop Assembly | MoF: 59.0 -> 71.2 | Improved |
| add probabilistic embeddings to VASOT | VASOT · Desktop Assembly | MoF: 67.9 -> 71.1 | Improved |
| add probabilistic embeddings to VASOT | VASOT · 50Salads (Eval) | MoF: 53.4 -> 62.2 | Improved |
| add probabilistic embedding (with MLP) to ASOT without GCN (ablation) | ASOT · Breakfast | F1: 35.7 -> 34.9 | Hurt |
| add probabilistic embedding (with MLP) to ASOT without GCN (ablation) | ASOT · Breakfast | MoF: 56.4 -> 59.2 | Improved |
| add GCN to probabilistic embedding (ablation) | PEOT · Breakfast | F1: 34.9 -> 40.5 | Improved |
| add probabilistic embedding to GCN (ablation) | PEOT · Desktop Assembly | MoF: 63.8 -> 71.2 | Improved |
| add probabilistic embedding to GCN (ablation) | PEOT · 50Salads (Eval) | MoF: 50.4 -> 64.9 | Improved |
| use GCN instead of TCN (ablation) | PEOT · Breakfast | F1: 38.9 -> 40.5 | Improved |
| use weighted adjacency instead of unweighted (ablation) | PEOT · Breakfast | MoF: 59.5 -> 60.7 | Improved |
| connect 3 frames instead of 5 (ablation) | PEOT · Breakfast | MoF: 58.9 -> 60.7 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CLOT — CLOT: Closed Loop Optimal Transport for Unsupervised Action Segmentation (ICCV 2025) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Unsupervised Action Segmentation by Joint Representation Learning and Online Clustering (CVPR 2022 · similarity 0.91)
- Temporally Consistent Unbalanced Optimal Transport for Unsupervised Action Segmentation (CVPR 2024 · similarity 0.90)
- Action Shuffle Alternating Learning for Unsupervised Action Segmentation (CVPR 2021 · similarity 0.90)
- Unsupervised Action Segmentation via Fast Learning of Semantically Consistent Actoms (AAAI 2024 · similarity 0.88)
- Hierarchical Vector Quantization for Unsupervised Action Segmentation (AAAI 2025 · similarity 0.87)
- Efficient and Effective Weakly-Supervised Action Segmentation via Action-Transition-Aware Boundary Alignment (CVPR 2024 · 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.
