Research map / Vision-language and multimodal / CLIP and zero-shot
DynaPURLS: Dynamic Refinement of Part-Aware Representations for Skeleton-Based Zero-Shot Action Recognition
Jingmin Zhu, Anqi Zhu, James Bailey, Jun Liu, Hossein Rahmani, Mohammed Bennamoun, Farid Boussaïd, Qiuhong Ke
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 inference-time dynamic refinement (ablation) | adaptive alignment model · NTU60 55/5 | ZSL accuracy: 79.23 -> 88.52 (+9.29) | Improved |
| add inference-time dynamic refinement (ablation) | adaptive alignment model · NTU60 48/12 | ZSL accuracy: 40.99 -> 71.80 (+30.81) | Improved |
| add inference-time dynamic refinement (ablation) | adaptive alignment model · NTU120 110/10 | ZSL accuracy: 71.95 -> 89.06 (+17.11) | Improved |
| add class-balanced memory bank over naive pseudo-labeling (ablation) | DynaPURLS refinement · NTU60 48/12 | ZSL accuracy: 51.60 -> 71.80 | Improved |
| naive pseudo-labeling refinement without memory bank (ablation) | PURLS baseline · NTU60 55/5 | ZSL accuracy: 79.22 -> 78.02 | Hurt |
| replace static body-part partition with adaptive attention partition (ablation) | PURLS-style alignment · NTU60 55/5 | ZSL accuracy: 76.46 -> 79.23 | Improved |
| replace global alignment with multi-granularity alignment (ablation) | Shift-GCN encoder, GPT-3 · NTU60 55/5 | ZSL accuracy: 64.69 -> 79.23 | Improved |
| replace global alignment with multi-granularity alignment (ablation) | PoseC3D encoder, GPT-3 · NTU60 55/5 | ZSL accuracy: 63.45 -> 81.14 | Improved |
| add temporal-interval alignment to body-part alignment, with refinement (ablation) | DynaPURLS · NTU60 48/12 | ZSL accuracy: 61.46 -> 71.80 | Improved |
| use GPT-4 instead of GPT-3 descriptors (ablation) | DynaPURLS · NTU60 55/5 | ZSL accuracy: 79.23 -> 81.53 | No clear effect |
| use random instead of identity initialization of refinement parameters (ablation) | DynaPURLS · NTU60 55/5 | ZSL accuracy: 88.52 -> 88.31 | No clear effect |
| adapt on full test distribution instead of seen only (ablation) | DynaPURLS GZSL · NTU60 48/12 | unseen accuracy: 36.92 -> 61.81 | Improved |
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.
- Neuron — Neuron: Learning Context-Aware Evolving Representations for Zero-Shot Skeleton Action Recognition (CVPR 2025) · better on 4 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
- Part-aware Unified Representation of Language and Skeleton for Zero-shot Action Recognition (CVPR 2024 · similarity 0.93)
- Neuron: Learning Context-Aware Evolving Representations for Zero-Shot Skeleton Action Recognition (CVPR 2025 · similarity 0.91)
- Boosting Skeleton-based Zero-Shot Action Recognition with Training-Free Test-Time Adaptation (NeurIPS 2025 · similarity 0.90)
- Semantic-guided Cross-Modal Prompt Learning for Skeleton-based Zero-shot Action Recognition (CVPR 2025 · similarity 0.89)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 2025 · similarity 0.88)
- Recovering Complete Actions for Cross-dataset Skeleton Action Recognition (NeurIPS 2024 · 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.
