Research map / Video / Action recognition
Unsupervised Video Understanding by Reconciliation of Posture Similarities
Timo Milbich, Miguel Bautista, Ekaterina Sutter, Bjorn Ommer
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 |
|---|---|---|---|
| use sequence-matching self-supervision instead of CliqueCNN | posture representation (scratch) · Olympic Sports | average AUC: 0.79 -> 0.83 | Improved |
| train with cross-sequence sequence matching instead of tuple/triplet methods | posture CNN from scratch · Olympic Sports | average AUC: 0.63 -> 0.83 | Improved |
| add ImageNet pre-trained initialization (ablation) | posture CNN · Olympic Sports | average AUC: 0.83 -> 0.83 | No clear effect |
| add ImageNet initialization (ablation) | posture CNN · LSP zero-shot | PCP@0.5 total: 43.0 -> 50.0 | Improved |
| use our ImageNet-initialized representation instead of CliqueCNN | posture CNN · LSP zero-shot | PCP@0.5 total: 43.5 -> 50.0 | Improved |
| use our scratch representation instead of Shuffle&Learn | posture CNN · LSP zero-shot | PCP@0.5 total: 30.0 -> 43.0 | Improved |
| initialize DeepPose with our unsupervised representation instead of random init | DeepPose · MPII Pose | PCKh@0.5 total: 65.4 -> 70.5 | Improved |
| initialize with ours instead of Shuffle&Learn | DeepPose · MPII Pose | PCKh@0.5 total: 69.3 -> 70.5 | Improved |
| combine our pretraining with ImageNet initialization | DeepPose · MPII Pose | PCKh@0.5 total: 78.0 -> 80.4 | Improved |
| use our scratch representation instead of ImageNet initialization | DeepPose · MPII Pose | PCKh@0.5 total: 78.0 -> 70.5 | Hurt |
Closest papers by meaning
- Unsupervised Learning of View-invariant Action Representations (NeurIPS 2018 · similarity 0.87)
- PREDICT & CLUSTER: Unsupervised Skeleton Based Action Recognition (CVPR 2020 · similarity 0.87)
- Recovering Complete Actions for Cross-dataset Skeleton Action Recognition (NeurIPS 2024 · similarity 0.85)
- Unsupervised Learning of Category-Level 3D Pose from Object-Centric Videos (CVPR 2024 · similarity 0.85)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 2025 · similarity 0.85)
- Unsupervised Action Segmentation via Fast Learning of Semantically Consistent Actoms (AAAI 2024 · similarity 0.85)
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.
