Research map / Video / Action recognition
SSAN: Separable Self-Attention Network for Video Representation Learning
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 |
|---|---|---|---|
| Replace ResNeXt-101 video features with SSA video representations | GRU video-text retrieval model · MSR-VTT | R@1: 12.1 -> 24.4 | Helped |
| Replace ResNeXt-101 video features with SSA video representations | GRU video-text retrieval model · MSR-VTT | R@5: 35.0 -> 49.3 | Helped |
| Replace ResNeXt-101 video features with SSA video representations | GRU video-text retrieval model · Youcook2 | R@1: 4.2 -> 5.5 | Helped |
| Add spatial attention (SA) to 2D+1D NL (ablation) | 2D+1D NL · Something-Something-V2 | Top-1 accuracy: 61.0 -> 61.5 | Helped |
| Add shared value embedding (SVE) (ablation) | 2D+1D NL with SA · Something-Something-V2 | Top-1 accuracy: 61.5 -> 61.4 | Hurt |
| Add temporal attention (TA) (ablation) | 2D+1D NL with SA and SVE · Something-Something-V2 | Top-1 accuracy: 61.4 -> 62.3 | Helped |
| Insert SSA in Res2 only (ablation) | ResNet-50 TSN · Something-Something-V2 | Top-1 accuracy: 61.2 | Helped |
| Insert SSA in Res3 only (ablation) | ResNet-50 TSN · Something-Something-V2 | Top-1 accuracy: 61.8 | Helped |
| Insert SSA in both Res2 and Res3 (ablation) | ResNet-50 TSN · Something-Something-V2 | Top-1 accuracy: 61.2 -> 62.3 | Helped |
| Insert SSA in later blocks Res4/Res5 instead of Res2/Res3 (ablation) | ResNet-50 TSN · Something-Something-V2 | Top-1 accuracy: around 0.5\%-0.7\% | Hurt |
Closest papers by meaning
- Self-Supervised Video Representation Learning via Latent Time Navigation (AAAI 2023 · similarity 0.88)
- Relational Self-Attention: What's Missing in Attention for Video Understanding (NeurIPS 2021 · similarity 0.88)
- Frame-Wise Action Representations for Long Videos via Sequence Contrastive Learning (CVPR 2022 · similarity 0.87)
- Self-Supervised Spatiotemporal Representation Learning by Exploiting Video Continuity (AAAI 2022 · similarity 0.86)
- StNet: Local and Global Spatial-Temporal Modeling for Action Recognition (AAAI 2019 · similarity 0.86)
- Self-Supervised Video Representation Learning by Context and Motion Decoupling (CVPR 2021 · 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.