Research map / Video / Action recognition
Action Recognition With Spatial-Temporal Discriminative Filter Banks
Brais Martinez, Davide Modolo, Yuanjun Xiong, Joseph Tighe
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 discriminative filter bank (ablation) | 2D baseline · Kinetics-400 | Top-1: 69.6 -> 70.9 | Improved |
| add discriminative filter bank (ablation) | 3D baseline · Kinetics-400 | Top-1: 66.8 -> 68.0 | Improved |
| add local branch on top of filters (ablation) | 2D model · Kinetics-400 | Top-1: 70.9 -> 71.2 | Improved |
| add local branch on top of filters (ablation) | 3D model · Kinetics-400 | Top-1: 68.0 -> 68.8 | Improved |
| add local branch on top of filters (ablation) | 2D model · Kinetics-400 | Top-5: 89.5 -> 89.3 | No clear effect |
| full approach versus baseline | 3D ResNet · Kinetics-400 meta-category Waxing | Top-1: +5.5 | Improved |
| full approach versus baseline | 3D ResNet · Kinetics-400 meta-category Watersports | Top-1: +0.5 | No clear effect |
| full approach versus baseline | 3D ResNet · Mopping floor confusion with cleaning floor | confused videos: 8 -> 4 | Improved |
| full approach versus baseline | 3D ResNet · Strumming guitar confusion with playing guitar | confused videos: 6 -> 13 | Hurt |
| full approach versus baseline | ResNet18 · Something-Something-V1 | Top-1: 42.3 -> 45.0 | Improved |
| deeper backbone ResNet152 instead of ResNet18 | Our approach · Something-Something-V1 | Top-1: 45.0 -> 53.4 | Improved |
| fine-grained components on TSN RGB stream | TSN one stream · Kinetics-400 (2D) | Top-1: 73.4 -> 74.3 | Improved |
Closest papers by meaning
- Deep Analysis of CNN-Based Spatio-Temporal Representations for Action Recognition (CVPR 2021 · similarity 0.91)
- VideoNet: A Large-Scale Dataset for Domain-Specific Action Recognition (CVPR 2026 · similarity 0.90)
- Sympathy for the Details: Dense Trajectories and Hybrid Classification Architectures for Action Recognition (ECCV 2016 · similarity 0.90)
- More Is Less: Learning Efficient Video Representations by Big-Little Network and Depthwise Temporal Aggregation (NeurIPS 2019 · similarity 0.90)
- Convolutional Two-Stream Network Fusion for Video Action Recognition (CVPR 2016 · similarity 0.89)
- Spatiotemporal Residual Networks for Video Action Recognition (NeurIPS 2016 · similarity 0.89)
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.
