Research map / Video / Action recognition
Real-Time Action Recognition With Enhanced Motion Vector CNNs
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 |
|---|---|---|---|
| replace optical flow with motion vectors and train from scratch | two-stream temporal CNN · UCF101 Split1 | accuracy: 81.2% -> 74.4% | Hurt |
| add teacher initialization from OF-CNN (ablation) | MV-CNN · UCF101 Split1 | accuracy: 74.4% -> 78.2% | Helped |
| add supervision transfer from OF-CNN (ablation) | MV-CNN · UCF101 Split1 | accuracy: 74.4% -> 77.5% | Helped |
| combine teacher initialization and supervision transfer (ablation) | MV-CNN · UCF101 Split1 | accuracy: 74.4% -> 79.3% | Helped |
| transfer optical-flow CNN knowledge to MV-CNN | MV-CNN · THUMOS14 | mAP: 29.8% -> 41.6% | Helped |
| use EMV-CNN rather than MV-CNN alongside the RGB CNN | RGB-plus-temporal two-stream system · THUMOS14 | mAP: 58.7% -> 61.5% | Helped |
| use 5 crops and mirroring instead of 1 crop (ablation) | EMV+RGB-CNN · UCF101 Split1 | accuracy: 85.7% -> 86.6% | Helped |
| use 5 crops and mirroring instead of 1 crop (ablation) | EMV+RGB-CNN · UCF101 Split1 | FPS: 390.7 -> 88.0 | Hurt |
| use 5 crops and mirroring instead of 1 crop (ablation) | EMV+RGB-CNN · THUMOS14 | mAP: 61.5% -> 61.2% | Hurt |
| vary the supervision-transfer temperature among 1, 2, and 3 (ablation) | EMV-CNN · UCF101 Split1 | accuracy: 79.2% -> 79.3% -> 78.9% | No clear effect |
Closest papers by meaning
- Optical Flow Guided Feature: A Fast and Robust Motion Representation for Video Action Recognition (CVPR 2018 · similarity 0.91)
- Representation Flow for Action Recognition (CVPR 2019 · similarity 0.89)
- MARS: Motion-Augmented RGB Stream for Action Recognition (CVPR 2019 · similarity 0.89)
- Convolutional Two-Stream Network Fusion for Video Action Recognition (CVPR 2016 · similarity 0.88)
- DMC-Net: Generating Discriminative Motion Cues for Fast Compressed Video Action Recognition (CVPR 2019 · similarity 0.88)
- Video Modeling With Correlation Networks (CVPR 2020 · 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.