Research map / Video / Action recognition
Video Representation Learning Using Discriminative Pooling
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 |
|---|---|---|---|
| Use SVMP instead of average pooling | Two-stream ResNet · HMDB-51 split-1 | accuracy: 63.8% -> 71.0% | Helped |
| Use SVMP instead of average pooling | Res-TCN · NTU-RGBD cross-subject | accuracy: 74.3% -> 78.5% | Helped |
| Use SVMP instead of average pooling | Res-TCN · NTU-RGBD cross-view | accuracy: 83.1% -> 86.4% | Helped |
| Apply SVMP instead of average pooling | Two-stream I3D+ · HMDB-51 (3 splits) | accuracy: 80.9% -> 81.3% | Helped |
| Combine linear and nonlinear SVMP (ablation) | VGG · HMDB-51 split-1 | accuracy: 63.8% -> 66.1% | Helped |
| Combine linear and nonlinear SVMP (ablation) | ResNet · HMDB-51 split-1 | accuracy: 69.5% -> 71.0% | Helped |
| Use nonlinear rather than linear SVMP (ablation) | VGG · HMDB-51 split-1 | accuracy: 63.8% -> 64.4% | Helped |
| Use nonlinear rather than linear SVMP (ablation) | ResNet · HMDB-51 split-1 | accuracy: 69.5% -> 69.8% | Helped |
| Use SVMP rather than average pooling for anticipation from the initial 1/5 | ResNet · HMDB-51 split-1, first 1/5 | accuracy: 48.6% -> 58.3% | Helped |
| Pool fc6 and pool5 features together (ablation) | VGG-16 · HMDB-51 split-1 | accuracy: 57.9% -> 63.8% | Helped |
| Use white-noise negative bags rather than the tested alternative negative bags (ablation) | SVMP on HMDB-51 split-1 · HMDB-51 split-1 | accuracy | Helped |
| Increase C beyond 10 (ablation) | SVMP · HMDB-51 | accuracy | Hurt |
Closest papers by meaning
- AdaScan: Adaptive Scan Pooling in Deep Convolutional Neural Networks for Human Action Recognition in Videos (CVPR 2017 · similarity 0.93)
- Learning End-to-end Video Classification with Rank-Pooling (ICML 2016 · similarity 0.91)
- Attentional Pooling for Action Recognition (NeurIPS 2017 · similarity 0.88)
- Dynamic Image Networks for Action Recognition (CVPR 2016 · similarity 0.88)
- Weakly Supervised Action Localization by Sparse Temporal Pooling Network (CVPR 2018 · similarity 0.88)
- DVANet: Disentangling View and Action Features for Multi-View Action Recognition (AAAI 2024 · similarity 0.87)
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.