Research map / Video / Action recognition
Delving into Details: Synopsis-to-Detail Networks for Video Recognition
Shuxian Liang, Xu Shen, Jianqiang Huang, Xian-Sheng Hua
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 DNet on top of SNet (ablation) | S2DNet · Mini-Kinetics | top-1 accuracy: 68.6 -> 74.7 | Improved |
| add SNet on top of DNet (ablation) | S2DNet · Mini-Kinetics | top-1 accuracy: 68.1 -> 74.7 | Improved |
| add learned location and scale parameters from SNet to DNet (ablation) | DNet · Mini-Kinetics | top-1 accuracy: 70.7 -> 73.3 | Improved |
| add contextual feature fusion to DNet (ablation) | DNet · Mini-Kinetics | top-1 accuracy: 70.7 -> 71.7 | Improved |
| combine sampling and fusion in DNet (ablation) | DNet · Mini-Kinetics | top-1 accuracy: 71.7 -> 74.7 | Improved |
| add spatial sampling only (ablation) | detail sampler · Mini-Kinetics | top-1 accuracy: 71.7 -> 72.7 | Improved |
| add temporal sampling only (ablation) | detail sampler · Mini-Kinetics | top-1 accuracy: 71.7 -> 74.0 | Improved |
| add learned space-time sampling over frozen center crops (ablation) | detail sampler · Mini-Kinetics | top-1 accuracy: 71.7 -> 74.7 | Improved |
| S2DNet over X3D-M | X3D-M · Kinetics-400 | top-1 accuracy: 76.0 -> 78.0 | Improved |
| S2DNet over X3D-XL | X3D-XL · Kinetics-400 | top-1 accuracy: 79.1 -> 80.6 | Improved |
| S2DNet over TSM | TSM R50 · Something-Something V2 | top-1 accuracy: 63.4 -> 66.4 | Improved |
| S2DNet vs TDN at lower cost | TDN R50 · Something-Something V2 | top-1 accuracy: 65.3 -> 66.4 | 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.
- AR-Net — AR-Net: Adaptive Frame Resolution for Efficient Action Recognition (ECCV 2020) · better on 1 of 1 shared result
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
- Action Detail Matters: Refining Video Recognition with Local Action Queries (CVPR 2025 · similarity 0.86)
- Deep Sequential Context Networks for Action Prediction (CVPR 2017 · similarity 0.86)
- I Know the Relationships: Zero-Shot Action Recognition via Two-Stream Graph Convolutional Networks and Knowledge Graphs (AAAI 2019 · similarity 0.85)
- FACT: Frame-Action Cross-Attention Temporal Modeling for Efficient Action Segmentation (CVPR 2024 · similarity 0.85)
- Intra- and Inter-Action Understanding via Temporal Action Parsing (CVPR 2020 · similarity 0.85)
- No Frame Left Behind: Full Video Action Recognition (CVPR 2021 · 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.
