Research map / Video / Action recognition
Temporal Pyramid Network for Action Recognition
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 |
|---|---|---|---|
| Add TPN to I3D-50 | I3D-50 · Kinetics-400 validation, 8×8 frames | Top-1 accuracy: 74.9 -> 76.1 (+1.2) | Helped |
| Add TPN to I3D-50 | I3D-50 · Kinetics-400 validation, 32×2 frames | Top-1 accuracy: 75.7 -> 77.7 (+2.0) | Helped |
| Add TPN to TSN-50 | TSN-50 · Kinetics-400 validation, 8 segments | Top-1 accuracy: 69.9 -> 73.5 | Helped |
| Add TPN to TSM-50 | TSM-50 · Something-Something V1 validation | Top-1 accuracy: 45.6 -> 49.0 | Helped |
| Add TPN to TSM-50 | TSM-50 · Something-Something V2 validation | Top-1 accuracy: 59.1 -> 62.0 | Helped |
| Add TPN to TSN | TSN · Epic-Kitchen validation | VERB@1: 48.2 -> 61.1 | Helped |
| Use res4 and res5 as feature sources rather than shallow res2/res3 stages (ablation) | TPN on I3D-50 · Kinetics-400 validation | Top-1 accuracy: 74.6 -> 76.1 | Helped |
| Add spatial convolutions to the auxiliary-head variant (ablation) | TPN on I3D-50 · Kinetics-400 validation | Top-1 accuracy: 74.6 -> 75.2 | Helped |
| Add temporal rate modulation after spatial modulation (ablation) | TPN on I3D-50 · Kinetics-400 validation | Top-1 accuracy: 75.2 -> 75.4 | Helped |
| Add information flow to the modulated TPN (ablation) | TPN on I3D-50 · Kinetics-400 validation | Top-1 accuracy: 75.4 -> 76.1 | Helped |
| Add the auxiliary classification head (ablation) | TPN on I3D-50 · Kinetics-400 validation | Top-1 accuracy: 75.9 -> 76.1 | Hurt |
| Use parallel information flow instead of isolation flow (ablation) | TPN on I3D-50 · Kinetics-400 validation | Top-1 accuracy: 75.4 -> 76.1 | Helped |
Closest papers by meaning
- T-C3D: Temporal Convolutional 3D Network for Real-Time Action Recognition (AAAI 2018 · similarity 0.91)
- Temporal Convolutional Networks for Action Segmentation and Detection (CVPR 2017 · similarity 0.90)
- TDN: Temporal Difference Networks for Efficient Action Recognition (CVPR 2021 · similarity 0.89)
- Memory-Augmented Temporal Dynamic Learning for Action Recognition (AAAI 2019 · similarity 0.89)
- MS-TCN: Multi-Stage Temporal Convolutional Network for Action Segmentation (CVPR 2019 · similarity 0.88)
- Spatiotemporal Pyramid Network for Video Action Recognition (CVPR 2017 · similarity 0.88)
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.