Research map / Video / Action recognition
CLIP-guided Prototype Modulating for Few-shot Action Recognition
Xiang Wang, Shiwei Zhang, Jun Cen, Changxin Gao, Yingya Zhang, Deli Zhao, Nong Sang
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 video-text contrastive objective (ablation) | OTAM baseline with CLIP-RN50 · Kinetics 5-way 1-shot | accuracy (%): 76.4 -> 80.4 | Improved |
| add video-text contrastive objective (ablation) | OTAM baseline with CLIP-RN50 · Kinetics 5-way 5-shot | accuracy (%): 88.9 -> 91.6 | Improved |
| add prototype modulation (ablation) | OTAM baseline with CLIP-RN50 · SSv2-Small 5-way 1-shot | accuracy (%): 39.9 -> 50.8 | Improved |
| add prototype modulation (ablation) | OTAM baseline with CLIP-RN50 · SSv2-Small 5-way 5-shot | accuracy (%): 51.9 -> 55.3 | Improved |
| add video-text objective on top of prototype modulation (ablation) | CLIP-FSAR · SSv2-Small 5-way 1-shot | accuracy (%): 50.8 -> 52.0 | Improved |
| add CLIP text features to temporal Transformer modulation (ablation) | Temporal Transformer-only · Kinetics 5-way 1-shot | accuracy (%): 77.9 -> 87.6 | Improved |
| use temporal Transformer instead of Bi-GRU (ablation) | CLIP-FSAR modulation · SSv2-Small 5-way 1-shot | accuracy (%): 47.1 -> 52.0 | Improved |
| share Transformer weights between support and query (ablation) | CLIP-FSAR · Kinetics 5-way 5-shot | accuracy (%): 84.7 -> 91.9 | Improved |
| increase temporal Transformer layers from 1 to 5 (ablation) | CLIP-FSAR · Kinetics 5-way 1-shot | accuracy (%): 87.6 -> 85.2 | Hurt |
| increase temporal Transformer layers from 1 to 3 (ablation) | CLIP-FSAR · SSv2-Small 5-way 1-shot | accuracy (%): 52.0 -> 52.4 | No clear effect |
| replace CLIP with ImageNet RN18 encoder plus CLIP-FSAR | OTAM · SSv2-Small 5-way 1-shot | accuracy (%): 35.8 -> 46.0 | Improved |
| plug CLIP-FSAR into TRX with CLIP-RN50 | TRX · Kinetics 5-way 1-shot | accuracy (%): 71.8 -> 86.7 | Improved |
Closest papers by meaning
- Frame Order Matters: A Temporal Sequence-Aware Model for Few-Shot Action Recognition (AAAI 2025 · similarity 0.87)
- Kronecker Mask and Interpretive Prompts are Language-Action Video Learners (ICLR 2025 · similarity 0.87)
- ImagineFSL: Self-Supervised Pretraining Matters on Imagined Base Set for VLM-based Few-shot Learning (CVPR 2025 · similarity 0.87)
- Text and Image Are Mutually Beneficial: Enhancing Training-Free Few-Shot Classification with CLIP (AAAI 2025 · similarity 0.87)
- BDC-CLIP: Brownian Distance Covariance for Adapting CLIP to Action Recognition (ICML 2025 · similarity 0.87)
- Building a Multi-modal Spatiotemporal Expert for Zero-shot Action Recognition with CLIP (AAAI 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.
