Research map / Video / Action recognition
Few-Shot Transformation of Common Actions Into Time and Space
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 |
|---|---|---|---|
| Use the few-shot transformer rather than the strongest listed baseline | Common action localization · Common-AVA and Common-UCF | frame-mAP: 21.6 -> 25.3; 23.2 -> 28.1; 59.4 -> 64.3; 61.3 -> 66.7 | Helped |
| Add encoder processing to support and query features (ablation) | Few-shot transformer · Common-AVA one-shot and five-shot | frame-mAP: 20.3 -> 25.6; 21.9 -> 28.1 | Helped |
| Add spatio-temporal positional encoding to support and query features (ablation) | Few-shot transformer · Common-AVA one-shot and five-shot | frame-mAP: 23.3 -> 25.6; 24.5 -> 28.1 | Helped |
| Add previous query-clip feature alignment (ablation) | Few-shot transformer · Common-AVA | frame-mAP | Helped |
| Increase the number or length of support videos (ablation) | Few-shot transformer · Common-AVA | frame-mAP | Helped |
| Add one support video showing a different action (ablation) | Few-shot transformer · Common-AVA five-shot | frame-mAP: 28.1 -> 26.8 | Hurt |
| Add one support video containing no action (ablation) | Few-shot transformer · Common-AVA five-shot | frame-mAP: 28.1 -> 26.3 | Hurt |
| Add eight noisy frames to each support video (ablation) | Few-shot transformer · Common-AVA five-shot | frame-mAP: 28.1 -> 24.5 | Hurt |
| Train on variable-length supports and evaluate on variable-length supports (ablation) | Few-shot transformer · Common-AVA five-shot | frame-mAP: 28.1 -> 26.1 | Hurt |
| Evaluate variable-length supports after training on 25-frame supports (ablation) | Few-shot transformer · Common-AVA five-shot | frame-mAP: 28.1 -> 25.0 | Hurt |
| Use the few-shot transformer rather than Yang et al. | Common action localization in time · Common-Instance and Common-Multi-instance | video-mAP: 53.1 -> 59.3; 56.5 -> 61.9; 42.1 -> 50.2; 43.9 -> 52.3 | Helped |
| Add the few-shot transformer mask-head rather than the adapted Siam et al. baseline | Common action localization per pixel · Common-A2D one-shot and five-shot | mIoU: 43.3 -> 50.6; 44.8 -> 52.5 | Helped |
Closest papers by meaning
- CDC: Convolutional-De-Convolutional Networks for Precise Temporal Action Localization in Untrimmed Videos (CVPR 2017 · similarity 0.90)
- Temporal Action Localization by Structured Maximal Sums (CVPR 2017 · similarity 0.89)
- Weakly-Supervised Temporal Action Localization by Inferring Salient Snippet-Feature (AAAI 2024 · similarity 0.88)
- End-to-End Spatio-Temporal Action Localisation with Video Transformers (CVPR 2024 · similarity 0.87)
- Realigning Confidence with Temporal Saliency Information for Point-Level Weakly-Supervised Temporal Action Localization (CVPR 2024 · similarity 0.87)
- Bridge the Gap: From Weak to Full Supervision for Temporal Action Localization with PseudoFormer (CVPR 2025 · 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.