Research map / Video / Action recognition
Otter: Mitigating Background Distractions of Wide-Angle Few-Shot Action Recognition with Enhanced RWKV
Wenbo Huang, Jinghui Zhang, Zhenghao Chen, Guang Li, Lei Zhang, Yang Cao, Fang Dong, Takahiro Ogawa, Miki Haseyama
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 CSM (compound segmentation module) (ablation) | Otter · SSv2 5-way 1-shot | accuracy (%): 54.6 -> 61.3 | Improved |
| add TRM (temporal reconstruction module) (ablation) | Otter · SSv2 5-way 1-shot | accuracy (%): 54.6 -> 59.5 | Improved |
| add TRM on top of CSM (ablation) | Otter · SSv2 5-way 1-shot | accuracy (%): 61.3 -> 64.7 | Improved |
| add CSM on top of TRM (ablation) | Otter · Kinetics 5-way 1-shot | accuracy (%): 87.8 -> 90.5 | Improved |
| use patch size 56 instead of 224 (ablation) | Otter · SSv2 5-way 1-shot | accuracy (%): 62.7 -> 64.7 | Improved |
| use patch size 56 instead of 28 (ablation) | Otter · SSv2 5-way 1-shot | accuracy (%): 64.1 -> 64.7 | Improved |
| use multi-scale patches {28,56} instead of single p=56 (ablation) | Otter · SSv2 5-way 1-shot | accuracy (%): 64.7 -> 64.2 | Hurt |
| add reversed scanning to ordered scanning in TRM (ablation) | Otter · SSv2 5-way 1-shot | accuracy (%): 63.2 -> 64.7 | Improved |
| add ordered scanning to reversed scanning in TRM (ablation) | Otter · SSv2 5-way 1-shot | accuracy (%): 60.6 -> 64.7 | Improved |
| use Otter instead of Manta on a wide-angle dataset | Few-shot action recognition · VideoBadminton (VB to VB, 1-shot) | accuracy (%): 64.1 -> 71.2 | Improved |
| use Otter instead of SOAP/Manta on UCF101 5-shot | ResNet-50 backbone · UCF101 5-way 5-shot | accuracy (%): 99.3 -> 99.2 | No clear effect |
| increase field of view level (outpainting and distortion) for Otter vs baselines | Otter · VideoBadminton 1-shot FoV levels 0-4 | accuracy (%): 71.2 -> 69.8 (Otter, Lv.0 to Lv.4) | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Manta — Manta: Enhancing Mamba for Few-Shot Action Recognition of Long Sub-Sequence (AAAI 2025) · better on 5 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- TAMT: Temporal-Aware Model Tuning for Cross-Domain Few-Shot Action Recognition (CVPR 2025 · similarity 0.87)
- Task-Specific Distance Correlation Matching for Few-Shot Action Recognition (AAAI 2026 · similarity 0.85)
- CLIP-guided Prototype Modulating for Few-shot Action Recognition (IJCV 2023 · similarity 0.85)
- Temporal Alignment-Free Video Matching for Few-shot Action Recognition (CVPR 2025 · similarity 0.84)
- Manta: Enhancing Mamba for Few-Shot Action Recognition of Long Sub-Sequence (AAAI 2025 · similarity 0.84)
- Building a Multi-modal Spatiotemporal Expert for Zero-shot Action Recognition with CLIP (AAAI 2025 · similarity 0.82)
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.
