Research map / Video / Action recognition
Temporal-Relational CrossTransformers for Few-Shot Action Recognition
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 ordered pairs and triples rather than single frames (ablation) | TRX · SSv2† 5-way 5-shot | accuracy: 53.3 -> 59.1 | Helped |
| Use ordered pairs and triples rather than single frames (ablation) | TRX · Kinetics 5-way 5-shot | accuracy: 85.2 -> 85.9 | Mixed |
| Combine pairs and triples (ablation) | TRX · SSv2† 5-way 5-shot | accuracy: 58.8 -> 59.1 | Helped |
| Add quadruples to pairs and triples (ablation) | TRX · Kinetics 5-way 5-shot | accuracy: 85.9 -> 85.3 | Hurt |
| Add quadruples to pairs and triples (ablation) | TRX · SSv2† 5-way 5-shot | accuracy: 59.1 -> 58.9 | Hurt |
| Reverse query tuple order at inference (ablation) | TRX Ω={2,3} · SSv2† 5-way 5-shot | accuracy: 59.1 -> 51.3 | Hurt |
| Reverse query tuple order at inference (ablation) | TRX Ω={2,3} · Kinetics 5-way 5-shot | accuracy: 85.9 -> 85.9 | No clear effect |
| Add positional encoding to tuple frame representations (ablation) | TRX Ω={2,3} · SSv2† 5-way 5-shot | accuracy: 58.5 -> 59.1 | Helped |
| Add positional encoding to tuple frame representations (ablation) | TRX Ω={2,3} · Kinetics 5-way 5-shot | accuracy: 85.5 -> 85.9 | Helped |
| Increase support examples from 1-shot to 5-shot (ablation) | TRX Ω={2,3} · SSv2† | accuracy: 42.0 -> 64.6 | Helped |
| Match query tuples against multiple support videos (ablation) | TRX · SSv2† 5-way few-shot | accuracy: +10.3% | Helped |
| Retain randomly selected tuples instead of exhaustive tuples (ablation) | TRX Ω={2,3} · SSv2† | accuracy | Mixed |
Closest papers by meaning
- Frame Order Matters: A Temporal Sequence-Aware Model for Few-Shot Action Recognition (AAAI 2025 · similarity 0.89)
- TA2N: Two-Stage Action Alignment Network for Few-Shot Action Recognition (AAAI 2022 · similarity 0.88)
- Temporal Alignment-Free Video Matching for Few-shot Action Recognition (CVPR 2025 · similarity 0.88)
- Adapting Short-Term Transformers for Action Detection in Untrimmed Videos (CVPR 2024 · similarity 0.87)
- FACT: Frame-Action Cross-Attention Temporal Modeling for Efficient Action Segmentation (CVPR 2024 · similarity 0.87)
- Learning Causal Domain-Invariant Temporal Dynamics for Few-Shot Action Recognition (ICML 2024 · 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.