Research map / Video / Action recognition
Spatio-Temporal Relation Modeling 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 |
|---|---|---|---|
| add spatio-temporal enrichment and query-class similarity classifier to TRX | TRX, ResNet-50 · SSv2 | accuracy: 64.6 -> 68.1 | Helped |
| add patch-level enrichment (PLE) (ablation) | Baseline TRM · SSv2 | accuracy: 62.1 -> 63.7 | Helped |
| add frame-level enrichment (FLE) (ablation) | Baseline TRM · SSv2 | accuracy: +3.2% | Helped |
| add PLE and FLE together (ablation) | Baseline TRM · SSv2 | accuracy: 62.1 -> 66.8 | Helped |
| add query-class similarity classification (ablation) | TRM with PLE and FLE · SSv2 | accuracy: 66.8 -> 68.1 | Helped |
| use joint spatio-temporal self-attention instead of separate PLE and FLE (ablation) | STRM · SSv2 | accuracy: 68.1 -> 64.7 | Hurt |
| use two layers of joint spatio-temporal self-attention (ablation) | STRM · SSv2 | accuracy: 68.1 -> 65.8 | Hurt |
| replace local PLE self-attention with MLP-Mixer and use self-attention for global FLE (ablation) | STRM · SSv2 | accuracy: 68.1 -> 64.1 | Hurt |
| use self-attention for both PLE and FLE (ablation) | STRM · SSv2 | accuracy: 68.1 -> 64.2 | Hurt |
| use MLP-Mixer for both PLE and FLE (ablation) | STRM · SSv2 | accuracy: 68.1 -> 65.0 | Hurt |
| use cardinality Ω={2,3,4} rather than Ω={2} (ablation) | STRM · Kinetics | accuracy: 86.5 -> 86.1 | Hurt |
| use cardinality Ω={2,3,4} rather than Ω={2} (ablation) | STRM · SSv2 | accuracy: 68.1 -> 68.1 | No clear effect |
Closest papers by meaning
- Revisiting the Spatial and Temporal Modeling for Few-Shot Action Recognition (AAAI 2023 · similarity 0.90)
- Searching for Better Spatio-temporal Alignment in Few-Shot Action Recognition (NeurIPS 2022 · similarity 0.88)
- Action Detail Matters: Refining Video Recognition with Local Action Queries (CVPR 2025 · similarity 0.87)
- Frame Order Matters: A Temporal Sequence-Aware Model for Few-Shot Action Recognition (AAAI 2025 · similarity 0.87)
- StNet: Local and Global Spatial-Temporal Modeling for Action Recognition (AAAI 2019 · similarity 0.86)
- Building a Multi-modal Spatiotemporal Expert for Zero-shot Action Recognition with CLIP (AAAI 2025 · 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.