Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
DASZL: Dynamic Action Signatures for Zero-shot Learning
Read the paperCheck 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 dynamic rather than static action signatures (ablation) | DASZL · JIGSAWS gesture classification | accuracy: 58.9% -> 69.7% | Helped |
| add the suturing-task grammar (ablation) | DASZL · JIGSAWS joint classification and segmentation | Edit Score: 32.7 -> 61.7 | Helped |
| add the suturing-task grammar (ablation) | DASZL · JIGSAWS joint classification and segmentation | frame-level accuracy: 48.5 -> 56.6 | Helped |
| use DASZL instead of the strongest listed inductive baseline | Olympic Sports ZSL (ID) · Olympic Sports | mean-per-class accuracy: 56.5 ± 6.6 -> 74.2 ± 9.9 | Helped |
| use DASZL instead of the strongest listed inductive baseline | UCF101 ZSL (ID) · UCF101 | mean-per-class accuracy: 48.0 ± 0.0 -> 48.9 ± 5.8 | Helped |
| use DASZL instead of the strongest listed transductive baseline | Olympic Sports ZSL (TD) · Olympic Sports | mean-per-class accuracy: 65.9 ± 8.1 -> 74.2 ± 9.9 | Helped |
| use DASZL instead of the strongest listed transductive baseline | UCF101 ZSL (TD) · UCF101 | mean-per-class accuracy: 41.6 ± 3.7 -> 48.9 ± 5.8 | Helped |
| use DASZL instead of the strongest listed transductive baseline | Olympic Sports GZSL (TD) · Olympic Sports | harmonic mean accuracy: 66.2 ± 6.3 -> 61.0 ± 4.9 | Hurt |
| use DASZL instead of the strongest listed transductive baseline | UCF101 GZSL (TD) · UCF101 | harmonic mean accuracy: 49.4 ± 2.4 -> 45.3 ± 2.8 | Hurt |
| add the binary in-vs-out-of-distribution auxiliary classifier (ablation) | DASZL · Olympic Sports and UCF101 GZSL | harmonic mean accuracy: 4.6 and 1.5 | Helped |
| compose dynamic action signatures using off-the-shelf object detectors rather than end-to-end supervised action models | DASZL · DIVA held-out scene 0000 | accuracy: 57.1 -> 85.0; 63.7 -> 85.0 | Helped |
| use dynamic action signatures rather than static word embeddings | DASZL versus E2E-K664 · DIVA held-out scene 0000 | accuracy: 50.0 -> 85.0 | Helped |
Closest papers by meaning
- ActionAtlas: A VideoQA Benchmark for Domain-specialized Action Recognition (NeurIPS 2024 · similarity 0.86)
- ZeroHAR: Sensor Context Augments Zero-Shot Wearable Action Recognition (AAAI 2025 · similarity 0.85)
- Beyond Image Classification: A Video Benchmark and Dual-Branch Hybrid Discrimination Framework for Compositional Zero-Shot Learning (CVPR 2025 · similarity 0.85)
- SeFAR: Semi-supervised Fine-grained Action Recognition with Temporal Perturbation and Learning Stabilization (AAAI 2025 · similarity 0.85)
- Boosting Skeleton-based Zero-Shot Action Recognition with Training-Free Test-Time Adaptation (NeurIPS 2025 · similarity 0.84)
- Building a Multi-modal Spatiotemporal Expert for Zero-shot Action Recognition with CLIP (AAAI 2025 · similarity 0.84)
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.