Research map / Video / Action recognition
Trokens: Semantic-Aware Relational Trajectory Tokens for Few-Shot Action Recognition
Pulkit Kumar, Shuaiyi Huang, Matthew Walmer, Sai Saketh Rambhatla, Abhinav Shrivastava
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 |
|---|---|---|---|
| add semantic-aware point sampling (ablation) | Trokens · SSV2 Small 1-shot | accuracy: 47.9 -> 49.9 | Improved |
| add semantic-aware point sampling (ablation) | Trokens · SSV2 Small 5-shot | accuracy: 64.4 -> 65.7 | Improved |
| add semantic-aware point sampling (ablation) | Trokens · SSV2 Full 1-shot | accuracy: 57.7 -> 58.6 | Improved |
| add semantic-aware point sampling (ablation) | Trokens · SSV2 Full 5-shot | accuracy: 74.6 -> 74.2 | No clear effect |
| add intra-motion module (HoD) after sampling (ablation) | Trokens · SSV2 Full 1-shot | accuracy: 58.6 -> 61.3 | Improved |
| add inter-motion module after sampling (ablation) | Trokens · SSV2 Full 1-shot | accuracy: 58.6 -> 60.6 | Improved |
| add inter-motion to sampling plus intra-motion (ablation) | Trokens · SSV2 Small 1-shot | accuracy: 51.8 -> 53.4 | Improved |
| add intra-motion to sampling plus inter-motion (ablation) | Trokens · SSV2 Small 1-shot | accuracy: 52.2 -> 53.4 | Improved |
| use HoD instead of displacement-only intra-motion features (ablation) | Trokens · SSV2 Full 1-shot | accuracy: 59.9 -> 61.3 | Improved |
| use HoD instead of displacement-only features (ablation) | Trokens · SSV2 Full 3-shot | accuracy: 71.6 -> 73.6 | Improved |
| increase TATs parameters from 11.8M to 18.1M to match model size (ablation) | TATs · SSV2 Full 1-shot | accuracy: 57.7 -> 59.6 | No clear effect |
| increase TATs parameters to 18.1M (ablation) | TATs · SSV2 Small 5-shot | accuracy: 64.4 -> 63.0 | Hurt |
Closest papers by meaning
- Trajectory-aligned Space-time Tokens for Few-shot Action Recognition (ECCV 2024 · similarity 0.91)
- Frame Order Matters: A Temporal Sequence-Aware Model for Few-Shot Action Recognition (AAAI 2025 · similarity 0.87)
- TrAction: Action Recognition with Sparse Trajectories (NeurIPS 2026 · similarity 0.87)
- Learning Causal Domain-Invariant Temporal Dynamics for Few-Shot Action Recognition (ICML 2024 · similarity 0.86)
- Temporal Alignment-Free Video Matching for Few-shot Action Recognition (CVPR 2025 · similarity 0.86)
- Disentangled Concepts Speak Louder Than Words: Explainable Video Action Recognition (NeurIPS 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.
