Research map / Video / Action recognition
Learning Comprehensive Motion Representation for 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 CME and SME to ResNet-50 + TIM (ablation) | ResNet-50 + TIM · Something-Something V1 | top-1 accuracy: 46.1% -> 51.3% | Helped |
| Add SME to ResNet-50 + TIM + CME (ablation) | ResNet-50 + TIM · Something-Something V1 | top-1 accuracy: 50.6% -> 51.3% | Helped |
| Add CME to ResNet-50 + TIM (ablation) | ResNet-50 + TIM · Something-Something V1 | top-1 accuracy: 46.1% -> 50.6% | Helped |
| Add SME to ResNet-50 + TIM (ablation) | ResNet-50 + TIM · Something-Something V1 | top-1 accuracy: 46.1% -> 50.6% | Helped |
| Add TIM to ResNet-50 + CME + SME (ablation) | ResNet-50 + CME + SME · Something-Something V1 | top-1 accuracy: 43.5% -> 51.3% | Helped |
| Set CME channel scaling factors r1=r2=1 instead of 8 (ablation) | CMR · Something-Something V1 | top-1 accuracy: 51.3% -> 47.3% | Hurt |
| Set CME channel scaling factors r1=r2=4 instead of 8 (ablation) | CMR · Something-Something V1 | top-1 accuracy: 51.3% -> 50.6% | Hurt |
| Set CME channel scaling factors r1=r2=16 instead of 8 (ablation) | CMR · Something-Something V1 | top-1 accuracy: 51.3% -> 50.3% | Hurt |
| Insert SME in every ResNet block instead of only the first block per stage (ablation) | CMR with CME in every block · Something-Something V1 | top-1 accuracy: 51.3% -> 51.0% | Hurt |
| Insert CME and SME only in each stage's first block (ablation) | CMR · Something-Something V1 | top-1 accuracy: 51.3% -> 50.4% | Hurt |
| Replace CME with MEM (ablation) | ResNet-50 + TIM motion enhancement · Something-Something V1 | top-1 accuracy: 47.4% -> 50.6% | Hurt |
| Replace CME with PEM (ablation) | ResNet-50 + TIM motion enhancement · Something-Something V1 | top-1 accuracy: 48.7% -> 50.6% | Hurt |
Closest papers by meaning
- ACTION-Net: Multipath Excitation for Action Recognition (CVPR 2021 · similarity 0.90)
- Deep Analysis of CNN-Based Spatio-Temporal Representations for Action Recognition (CVPR 2021 · similarity 0.88)
- Video Modeling With Correlation Networks (CVPR 2020 · similarity 0.87)
- Real-Time Action Recognition With Enhanced Motion Vector CNNs (CVPR 2016 · similarity 0.87)
- Convolutional Two-Stream Network Fusion for Video Action Recognition (CVPR 2016 · similarity 0.86)
- Prompt-guided Disentangled Representation for Action Recognition (NeurIPS 2025 · 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.