Research map / Label-efficient and robust learning / Semi-supervised learning
How Do You Do It? Fine-Grained Action Understanding With Pseudo-Adverbs
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 multi-adverb pseudo-labeling rather than a single pseudo-label per clip (ablation) | PseudoAdverbs · VATEX Adverbs, seen compositions | accuracy | Helped |
| Set the maximum pseudo-label count to k=7 (ablation) | PseudoAdverbs · VATEX Adverbs, seen compositions | accuracy | Helped |
| Allow very many pseudo-labels per clip (k≥13) (ablation) | PseudoAdverbs · VATEX Adverbs, seen compositions | accuracy | Hurt |
| Select pseudo-labels by antonym-relative maximum difference rather than closest embedding (ablation) | PseudoAdverbs · VATEX Adverbs, pseudo-label selection | accuracy: 61.7 -> 63.9 | Helped |
| Use per-adverb adaptive rather than fixed thresholds (ablation) | PseudoAdverbs · VATEX Adverbs, thresholding | accuracy: 61.4 -> 63.9 | Helped |
| Remove thresholding (ablation) | PseudoAdverbs · VATEX Adverbs, thresholding | accuracy: 63.9 -> 61.1 | Hurt |
| Add action-only videos for semi-supervised training (ablation) | PseudoAdverbs · VATEX Adverbs, seen compositions | accuracy | Helped |
| Use the proposed method instead of supervised-only training | VATEX Adverbs · Unseen compositions | accuracy: 52.2 -> 56.1 | Helped |
| Use the proposed method instead of source-only training | VATEX-to-MSR-VTT transfer · MSR-VTT Adverbs, unseen domain | accuracy: 62.9 -> 65.0 | Helped |
| Use the proposed method instead of source-only training | VATEX-to-ActivityNet transfer · ActivityNet Adverbs, unseen domain | accuracy: 67.2 -> 66.6 | Hurt |
| Combine RGB and optical-flow features rather than use flow alone (ablation) | PseudoAdverbs · VATEX Adverbs, modality comparison | accuracy: 64.5 -> 63.9 | Hurt |
| Increase the adaptive-threshold smoothing factor into 0.04≤λ≤0.14 (ablation) | PseudoAdverbs · VATEX Adverbs, seen compositions | accuracy | Helped |
Closest papers by meaning
- SeFAR: Semi-supervised Fine-grained Action Recognition with Temporal Perturbation and Learning Stabilization (AAAI 2025 · similarity 0.87)
- Storyboard-guided Alignment for Fine-grained Video Action Recognition (NeurIPS 2025 · similarity 0.86)
- ActionAtlas: A VideoQA Benchmark for Domain-specialized Action Recognition (NeurIPS 2024 · similarity 0.84)
- Open-Vocabulary Video Relation Extraction (AAAI 2024 · similarity 0.83)
- Kronecker Mask and Interpretive Prompts are Language-Action Video Learners (ICLR 2025 · similarity 0.83)
- SRTube: Video-Language Pre-Training with Action-Centric Video Tube Features and Semantic Role Labeling (CVPR 2024 · similarity 0.83)
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.