Research map / Video / Action recognition
Optical Flow Guided Feature: A Fast and Robust Motion Representation 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 OFF(RGB) to the RGB stream | RGB action recognition on UCF-101 · UCF-101 (Split1) | accuracy: 85.5% -> 90.0% | Helped |
| add OFF(RGB) to RGB + RGB Diff | RGB action recognition on UCF-101 · UCF-101 (Split1) | accuracy: 90.7% -> 92.0% | Helped |
| add OFF(RGB Diff) to RGB + OFF(RGB) + RGB Diff | RGB action recognition on UCF-101 · UCF-101 (Split1) | accuracy: 92.0% -> 93.0% | Helped |
| add OFF(RGB) to RGB + Flow | RGB and optical-flow action recognition · UCF-101 (Split1) | accuracy: 93.5% -> 95.1% | Helped |
| add OFF(Flow) to RGB + OFF(RGB) + Flow | RGB and optical-flow action recognition · UCF-101 (Split1) | accuracy: 95.1% -> 95.5% | Helped |
| add OFF(RGB) and OFF(RGB Diff) to RGB + RGB Diff | RGB-only fast action recognition · UCF-101 (3 splits) | accuracy: 85.5% -> 93.3% | Helped |
| add OFF(RGB Diff) to the final fused system | Ours (OFF) on UCF-101 and HMDB-51 · UCF-101 and HMDB-51 (3 splits) | accuracy: 0.3% | Helped |
| remove the OFF unit while retaining the hypercolumn structure (ablation) | RGB + OFF(RGB) on UCF-101 · UCF-101 (Split1) | accuracy: 86.0% -> 90.0% | Hurt |
| fine-tune the complete network end to end after the two training stages (ablation) | OFF network · UCF-101 and HMDB-51 | accuracy | No clear effect |
| add OFF(RGB) to the RGB stream | RGB-only action recognition on UCF-101 · UCF-101 (3 splits) | speed: 680 -> 450 | Hurt |
Closest papers by meaning
- Real-Time Action Recognition With Enhanced Motion Vector CNNs (CVPR 2016 · similarity 0.91)
- Representation Flow for Action Recognition (CVPR 2019 · similarity 0.91)
- Scene Flow to Action Map: A New Representation for RGB-D Based Action Recognition With Convolutional Neural Networks (CVPR 2017 · similarity 0.90)
- DMC-Net: Generating Discriminative Motion Cues for Fast Compressed Video Action Recognition (CVPR 2019 · similarity 0.89)
- MARS: Motion-Augmented RGB Stream for Action Recognition (CVPR 2019 · similarity 0.89)
- FlowFeat: Pixel-Dense Embedding of Motion Profiles (NeurIPS 2025 · similarity 0.86)
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.