Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Rethinking Zero-Shot Action Recognition: Learning from Latent Atomic Actions
Yijun Qian, Lijun Yu, Wenhe Liu, Alexander G. Hauptmann
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 group-to-group alignment of video segments and verb phrases (ablation) | JigsawNet · KineticsZSAR (3 splits) | rank-1 accuracy: 41.7 -> 45.9 | Improved |
| add group-to-group alignment (ablation) | JigsawNet · KineticsZSAR (3 splits) | rank-5 accuracy: 73.4 -> 78.8 | Improved |
| add consistency loss (ablation) | JigsawNet · KineticsZSAR (3 splits) | rank-1 accuracy: 43.2 -> 45.9 | Improved |
| add consistency loss (ablation) | JigsawNet · KineticsZSAR (3 splits) | rank-5 accuracy: 76.2 -> 78.8 | Improved |
| optimize consistency loss on aggregated instead of spatio-temporal features (ablation) | JigsawNet · KineticsZSAR (3 splits) | rank-1 accuracy: 45.1 -> 45.9 | Improved |
| add GE module (ablation) | JigsawNet · KineticsZSAR (3 splits) | rank-1 accuracy: 42.7 -> 45.9 | Improved |
| add GE module (ablation) | JigsawNet · KineticsZSAR (3 splits) | rank-5 accuracy: 72.9 -> 78.8 | Improved |
| deeper backbone R(2+1)D-34 instead of R(2+1)D-18 (ablation) | JigsawNet · KineticsZSAR (3 splits) | rank-1 accuracy: 42.9 -> 45.9 | Improved |
| deeper backbone R(2+1)D-34 instead of R(2+1)D-18 (ablation) | JigsawNet · KineticsZSAR (3 splits) | rank-5 accuracy: 73.0 -> 78.8 | Improved |
| TSM ResNet50 backbone vs R(2+1)D-34 (ablation) | JigsawNet · KineticsZSAR (3 splits) | rank-1 accuracy: 45.3 vs 45.9 | No clear effect |
| JigsawNet vs ER with same TSM extractor and inputs | JigsawNet · KineticsZSAR (3 splits) | rank-1 accuracy: 3.2 gains | Improved |
| JigsawNet vs ER | JigsawNet · HMDB51 / UCF101 | rank-1 accuracy: 3.4 and 4.2 gains | Improved |
Closest papers by meaning
- Reformulating Zero-shot Action Recognition for Multi-label Actions (NeurIPS 2021 · similarity 0.91)
- Elaborative Rehearsal for Zero-Shot Action Recognition (ICCV 2021 · similarity 0.90)
- Building a Multi-modal Spatiotemporal Expert for Zero-shot Action Recognition with CLIP (AAAI 2025 · similarity 0.88)
- Storyboard-guided Alignment for Fine-grained Video Action Recognition (NeurIPS 2025 · similarity 0.87)
- SeFAR: Semi-supervised Fine-grained Action Recognition with Temporal Perturbation and Learning Stabilization (AAAI 2025 · similarity 0.87)
- Text-Infused Attention and Foreground-Aware Modeling for Zero-Shot Temporal Action Detection (NeurIPS 2024 · similarity 0.86)
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.
