Research map / Video / Action recognition
Action Knowledge Transfer for Action Prediction with Partial Videos
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 two-stage action knowledge transfer from full videos | 3D ResNext-101 features · HMDB-51, ratio 0.1 | accuracy (%): 38.7 -> 43.5 | Helped |
| add two-stage action knowledge transfer from full videos | 3D ResNext-101 features · UCF-101, ratio 0.1 | accuracy (%): 76.3 -> 80.0 | Helped |
| add knowledge transfer on full videos | 3D ResNext-101 features · HMDB-51, ratio 1.0 | accuracy (%): 61.4 -> 61.8 | No clear effect |
| add discriminative classifier transfer to embedding transfer (ablation) | Embedding-Only · HMDB-51 split-1, ratio 0.1 | accuracy (%): 46.0 -> 46.8 | Helped |
| add embedding transfer to discriminative transfer (ablation) | Discriminative-Only · HMDB-51 split-1, ratio 0.1 | accuracy (%): 44.6 -> 46.8 | Helped |
| add AM Softmax for embedding learning (ablation) | Our method · HMDB-51 split-1, ratio 0.1 | accuracy (%): 44.8 -> 46.8 | Helped |
| use separate projection weights per progress level instead of shared (ablation) | Our method · HMDB-51 split-1, ratio 0.1 | accuracy (%): 44.6 -> 46.8 | Helped |
| weight progress levels by learned estimation instead of uniform scores (ablation) | Our method · HMDB-51 split-1, ratio 0.1 | accuracy (%): 44.4 -> 46.8 | Helped |
| use ground-truth progress level instead of estimated (ablation) | Our method · HMDB-51 split-1, ratio 0.1 | accuracy (%): 46.8 -> 46.5 | No clear effect |
| increase beta above 20000 (ablation) | Our method · HMDB-51 | mean accuracy | Hurt |
Closest papers by meaning
- Deep Sequential Context Networks for Action Prediction (CVPR 2017 · similarity 0.89)
- Progressive Teacher-Student Learning for Early Action Prediction (CVPR 2019 · similarity 0.88)
- Large-Scale Weakly-Supervised Pre-Training for Video Action Recognition (CVPR 2019 · similarity 0.87)
- Modular Action Concept Grounding in Semantic Video Prediction (CVPR 2022 · similarity 0.86)
- Unsupervised Action Segmentation via Fast Learning of Semantically Consistent Actoms (AAAI 2024 · similarity 0.85)
- Action Detail Matters: Refining Video Recognition with Local Action Queries (CVPR 2025 · 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.