Research map / Video / Action recognition
Temporal Bilinear Networks for Video Action 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 |
|---|---|---|---|
| Add TBN temporal bilinear blocks to C2D | ResNet-18, Kinetics · Kinetics validation | Top-1 accuracy: 61.1→65.0 | Helped |
| Add TBN temporal bilinear blocks to C2D | ResNet-34, Kinetics · Kinetics validation | Top-1 accuracy: 65.4→69.5 | Helped |
| Add TBN temporal bilinear blocks to C2D | ResNet-50, Kinetics · Kinetics validation | Top-1 accuracy: 66.9→70.1 | Helped |
| Add TBN to C2D | ResNet-18, ImageNet-pretrained transfer · UCF101 | Top-1 accuracy: 76.9 -> 77.8 | Helped |
| Add TBN to C2D | ResNet-18, ImageNet-pretrained transfer · HMDB51 | Top-1 accuracy: 41.2 -> 42.7 | Helped |
| Add TB blocks at res_1 rather than using C2D (ablation) | WTBN ResNet-18 · Mini-Kinetics-200 | Top-1 accuracy: 64.3 -> 68.3 | Helped |
| Add TB blocks at res_2 rather than using C2D (ablation) | WTBN ResNet-18 · Mini-Kinetics-200 | Top-1 accuracy: 64.3 -> 69.0 | Helped |
| Add TB blocks at res_3 rather than using C2D (ablation) | WTBN ResNet-18 · Mini-Kinetics-200 | Top-1 accuracy: 64.3 -> 69.2 | Helped |
| Add TB blocks only at res_4 (ablation) | WTBN ResNet-18 · Mini-Kinetics-200 | Top-1 accuracy: 64.3 -> 67.9 | Helped |
| Increase WTBN depth from 2 to 6 TB blocks (ablation) | WTBN ResNet-18 · Mini-Kinetics-200 | Top-1 accuracy: 67.9 -> 69.5 | Helped |
| Increase DTBN depth from 2 to 6 TB blocks (ablation) | DTBN ResNet-18 · Mini-Kinetics-200 | Top-1 accuracy: 66.9 -> 69.0 | Helped |
| Use the bottleneck design (ablation) | TBN with 4 TB blocks · Mini-Kinetics-200 | Top-1 accuracy: 66.6 -> 68.8 | Helped |
Closest papers by meaning
- Local Temporal Bilinear Pooling for Fine-Grained Action Parsing (CVPR 2019 · similarity 0.87)
- Spatiotemporal Pyramid Network for Video Action Recognition (CVPR 2017 · similarity 0.87)
- Temporal Convolutional Networks for Action Segmentation and Detection (CVPR 2017 · similarity 0.86)
- Temporal Pyramid Network for Action Recognition (CVPR 2020 · similarity 0.86)
- Trajectory Convolution for Action Recognition (NeurIPS 2018 · similarity 0.85)
- TEA: Temporal Excitation and Aggregation for Action Recognition (CVPR 2020 · 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.