Research map / Video / Action recognition
Improving Human Action Recognition by Non-Action Classification
Check 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 |
|---|---|---|---|
| Remove non-action shots using ground-truth labels | Dense Trajectories + Fisher Vector + LSSVM action recognition · ActionThread | mean AP: 45.3 -> 59.0 | Helped |
| Weight video segments using a generic non-action classifier | ActionThread action recognition with DTD · ActionThread | mean AP: 45.3 -> 52.1 | Helped |
| Use a generic rather than action-specific non-action classifier | ActionThread action recognition with DTD · ActionThread | mean AP: 49.3 -> 52.1 | Helped |
| Add CNN features to DTD (ablation) | Generic non-action weighted action recognition · ActionThread | mean AP: 52.1 -> 55.0 | Helped |
| Incorporate non-action scores into VideoDarwin | VideoDarwin action recognition · ActionThread | mean AP: 47.4 -> 52.2 | Helped |
| Combine DTD with spatial and temporal ConvNet features (ablation) | Non-action shot classifier · ActionThread non-action test set | AP: 84.7% -> 86.1% | Helped |
| Limit removals to at most 1 shot per video (ablation) | Generic non-action classifier · ActionThread non-action test set | AP: 86.1% -> 92.6% | Helped |
| Apply generic non-action pruning | DTD action recognition · Hollywood2 | mean AP: 64.2 -> 67.5 | Helped |
| Apply generic non-action pruning | DTD + CNN action recognition · Hollywood2 | mean AP: 67.2 -> 71.0 | Helped |
| Apply the ActionThread-trained generic non-action classifier | DTD action recognition · Six unseen actions | mean AP: 66.5 -> 70.7 | Helped |
| Apply the ActionThread-trained generic non-action classifier | DTD + CNN action recognition · Six unseen actions | mean AP: 66.9 -> 70.8 | Helped |
| Apply generic non-action pruning | DTD action recognition · Unseen actions: Drink | AP: 68.9 -> 68.6 | Hurt |
Closest papers by meaning
- Pulling Actions out of Context: Explicit Separation for Effective Combination (CVPR 2018 · similarity 0.88)
- Exploiting Motion Information from Unlabeled Videos for Static Image Action Recognition (AAAI 2020 · similarity 0.85)
- Learning Transferable Self-Attentive Representations for Action Recognition in Untrimmed Videos with Weak Supervision (AAAI 2019 · similarity 0.85)
- Action-Agnostic Point-Level Supervision for Temporal Action Detection (AAAI 2025 · similarity 0.84)
- Action Detail Matters: Refining Video Recognition with Local Action Queries (CVPR 2025 · similarity 0.83)
- ActionAtlas: A VideoQA Benchmark for Domain-specialized Action Recognition (NeurIPS 2024 · similarity 0.83)
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.