Research map / Video / Action recognition
ACTION-Net: Multipath Excitation 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 ACTION to ResNet-50 | TSM · Something-Something V2, 8 frames | Top-1 accuracy: 58.7 -> 62.5 | Helped |
| Add ACTION to ResNet-50 | TSM · Jester, 8 frames | Top-1 accuracy: 94.4 -> 97.1 | Helped |
| Add ACTION to ResNet-50 | TSM · EgoGesture, 8 frames | Top-1 accuracy: 92.1 -> 94.2 | Helped |
| Add ACTION to BNIception | TSM · Something-Something V2, Jester, and EgoGesture | Top-1 accuracy: 53.5 -> 56.7; 94.6 -> 96.6; 92.3 -> 93.2 | Helped |
| Add ACTION to MobileNet V2 | TSM · Something-Something V2, Jester, and EgoGesture | Top-1 accuracy: 54.9 -> 58.5; 95.0 -> 96.7; 92.4 -> 93.5 | Helped |
| Use the STE path (ablation) | ResNet-50 on EgoGesture · EgoGesture, 8 frames | Top-1 accuracy: 83.1 -> 93.8 | Helped |
| Use the CE path (ablation) | ResNet-50 on EgoGesture · EgoGesture, 8 frames | Top-1 accuracy: 83.1 -> 93.8 | Helped |
| Use the ME path (ablation) | ResNet-50 on EgoGesture · EgoGesture, 8 frames | Top-1 accuracy: 83.1 -> 93.9 | Helped |
| Combine STE, CE, and ME (ablation) | ResNet-50 on EgoGesture · EgoGesture, 8 frames | Top-1 accuracy: 92.1 -> 94.2 | Helped |
| Add ACTION to more residual stages (ablation) | ResNet-50 · EgoGesture, 8 frames | Top-1 accuracy: 92.3 -> 92.9 -> 93.1 -> 94.2 | Helped |
| Place ACTION at the start of the residual block rather than Loc 2, 3, or 4 (ablation) | ResNet-50 · EgoGesture, 8 segments | Top-1 accuracy: 94.0 -> 94.2; 93.8 -> 94.2; 94.0 -> 94.2 | Helped |
| Place ACTION at the start of the Inception block rather than Loc 2 (ablation) | BNInception · EgoGesture, 8 segments | Top-1 accuracy: 92.6 -> 93.2 | Helped |
Closest papers by meaning
- Deep Analysis of CNN-Based Spatio-Temporal Representations for Action Recognition (CVPR 2021 · similarity 0.91)
- Learning Comprehensive Motion Representation for Action Recognition (AAAI 2021 · similarity 0.90)
- Spatiotemporal Residual Networks for Video Action Recognition (NeurIPS 2016 · similarity 0.90)
- A Closer Look at Spatiotemporal Convolutions for Action Recognition (CVPR 2018 · similarity 0.90)
- Video Modeling With Correlation Networks (CVPR 2020 · similarity 0.89)
- Spatiotemporal Pyramid Network for Video Action Recognition (CVPR 2017 · 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.