Research map / Video / Action recognition
Watch-n-Patch: Unsupervised Learning of Actions and Relations
Chenxia Wu, Jiemi Zhang, Ozan Şener, Bart Selman, Silvio Savarese, Ashutosh Saxena
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 object-topics modeling to action-only CaTM (ablation) | CaTM · Watch-n-Patch office | PA-Acc (%): 41.5 -> 46.2 | Improved |
| add object-topics modeling to action-only CaTM (ablation) | CaTM · Watch-n-Patch kitchen | PA-Acc (%): 20.5 -> 24.4 | Improved |
| use skeleton and RGB-D features instead of DTF RGB features (ablation) | CaTM-AO · Watch-n-Patch office | PA-Acc (%): 36.2 -> 46.2 | Improved |
| use skeleton and RGB-D features instead of DTF RGB features (ablation) | HMM · Watch-n-Patch office | PA-Acc (%): 23.6 -> 33.3 | Improved |
| add action correlation priors to temporal-relation topic model (ablation) | TM-RT · Watch-n-Patch office | PA-Acc (%): 39.5 -> 46.2 | Improved |
| add long-range temporal relations to topic model (ablation) | TM · Watch-n-Patch office | Seg-Acc offline (%): 9.3 -> 30.8 | Improved |
| add topic correlation to TM (ablation) | TM · Watch-n-Patch office | PA-Acc (%): 13.3 -> 13.3 | No clear effect |
| add object modeling to HMM (ablation) | HMM · Watch-n-Patch kitchen | PA-Acc (%): 7.4 -> 12.4 | Improved |
| add object modeling to TM (ablation) | TM · Watch-n-Patch office | Seg-Acc offline (%): 9.3 -> 9.8 | No clear effect |
| CaTM-AO versus HMM-O in robot trials | Watch-Bot · Robotic experiment | success rate (%): 37.5 -> 62.5 | Improved |
| increase number of topics beyond ground-truth class count (ablation) | CaTM · Watch-n-Patch office | Seg-Acc/AP | Mixed |
Closest papers by meaning
- Unsupervised Learning of Action Classes With Continuous Temporal Embedding (CVPR 2019 · similarity 0.88)
- Latent Dirichlet Allocation for Unsupervised Activity Analysis on an Autonomous Mobile Robot (AAAI 2017 · similarity 0.87)
- What Actions Are Needed for Understanding Human Actions in Videos? (ICCV 2017 · similarity 0.87)
- Unsupervised Action Segmentation via Fast Learning of Semantically Consistent Actoms (AAAI 2024 · similarity 0.87)
- Recovering Complete Actions for Cross-dataset Skeleton Action Recognition (NeurIPS 2024 · similarity 0.86)
- HopaDIFF: Holistic-Partial Aware Fourier Conditioned Diffusion for Referring Human Action Segmentation in Multi-Person Scenarios (NeurIPS 2025 · similarity 0.86)
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.
