Research map / Video / Action recognition
Look More but Care Less in Video Recognition
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 AFNet with fewer selected frames rather than TSN | ResNet50 on Something-Something V1 · Something-Something V1 | Top-1 accuracy: 18.6$\%$ -> 27.7$\% | Helped |
| use AFNet with fewer selected frames rather than TSN | ResNet50 on Jester · Jester | Top-1 accuracy: 83.5$\%$ -> 89.2$\% | Helped |
| process 12 frames using AFNet-TSM rather than 8 using TSM | AF-ResNet101 on Something-Something V1 · Something-Something V1 | Top-1 accuracy: 45.6$\%$ -> 50.1$\% | Helped |
| use two branches rather than a single branch (ablation) | AFNet on ActivityNet · ActivityNet | mAP: 46.8$\%$ -> 72.2$\% | Helped |
| use the navigation module instead of random temporal sampling (ablation) | Two-branch AFNet on ActivityNet · ActivityNet | mAP: 62.8$\%$ -> 72.2$\% | Helped |
| use the navigation module instead of uniform temporal sampling (ablation) | Two-branch AFNet on ActivityNet · ActivityNet | mAP: 59.0$\%$ -> 72.2$\% | Helped |
| use the navigation module instead of normal temporal sampling (ablation) | Two-branch AFNet on ActivityNet · ActivityNet | mAP: 61.5$\%$ -> 72.2$\% | Helped |
| use spatial navigation instead of random spatial sampling (ablation) | Two-branch AFNet on ActivityNet · ActivityNet | mAP: 69.2$\%$ -> 72.0$\% | Helped |
| use spatial navigation instead of center cropping (ablation) | Two-branch AFNet on ActivityNet · ActivityNet | mAP: 68.9$\%$ -> 72.0$\% | Helped |
| replace dynamic feature fusion with addition (ablation) | AFNet on ActivityNet · ActivityNet | mAP: 74.3$\%$ -> 73.5$\% | Hurt |
| replace exponential temperature decay with cosine decay (ablation) | AFNet on ActivityNet · ActivityNet | mAP: 74.3$\%$ -> 74.1$\% | Hurt |
| replace exponential temperature decay with linear decay (ablation) | AFNet on ActivityNet · ActivityNet | mAP: 74.3$\%$ -> 73.8$\% | Hurt |
Closest papers by meaning
- More Is Less: Learning Efficient Video Representations by Big-Little Network and Depthwise Temporal Aggregation (NeurIPS 2019 · similarity 0.89)
- No Frame Left Behind: Full Video Action Recognition (CVPR 2021 · similarity 0.87)
- Deep Alternative Neural Network: Exploring Contexts as Early as Possible for Action Recognition (NeurIPS 2016 · similarity 0.87)
- Convolutional Two-Stream Network Fusion for Video Action Recognition (CVPR 2016 · similarity 0.87)
- Action Detail Matters: Refining Video Recognition with Local Action Queries (CVPR 2025 · similarity 0.85)
- DVANet: Disentangling View and Action Features for Multi-View Action Recognition (AAAI 2024 · similarity 0.85)
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.