Research map / Video / Action recognition
Discriminative Hierarchical Rank Pooling for Activity 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 |
|---|---|---|---|
| Use hierarchical rank pooling instead of recursive rank pooling | VGG-16 fc6 video encoding · Hollywood2 | mAP: 52.5 -> 56.8 | Helped |
| Use hierarchical rank pooling instead of recursive rank pooling | VGG-16 fc6 video encoding · HMDB51 | accuracy: 45.8 -> 47.5 | Helped |
| Use hierarchical rank pooling instead of recursive rank pooling | VGG-16 fc6 video encoding · UCF101 | accuracy: 75.6 -> 78.8 | Helped |
| Combine rank-pooled trajectory features with hierarchical rank-pooled CNN features | Activity recognition · Hollywood2 | mAP: 68.5 -> 74.1 | Helped |
| Combine rank-pooled trajectory features with hierarchical rank-pooled CNN features | Activity recognition · HMDB51 | accuracy: 60.0 -> 65.0 | Helped |
| Combine rank-pooled trajectory features with hierarchical rank-pooled CNN features | Activity recognition · UCF101 | accuracy: 86.5 -> 90.7 | Helped |
| Jointly learn the final-layer non-linear transform and classifier | Hierarchical rank pooling · UCF101 | accuracy: 78.8 -> 81.4 | Helped |
| Use SER instead of SSR for rank pooling (ablation) | Rank pooling · Hollywood2 | mAP: 44.2 -> 51.0 | Helped |
| Use SER instead of SSR for hierarchical rank pooling (ablation) | Hierarchical rank pooling · Hollywood2 | mAP: 50.5 -> 56.8 | Helped |
| Use SER instead of SSR (ablation) | Rank pooling on CNN activations · Hollywood2 | mAP: 44.2 -> 51.0 | Helped |
| Use SER instead of SSR (ablation) | Hierarchical rank pooling on CNN activations · Hollywood2 | mAP: 50.5 -> 56.8 | Helped |
| Use Chi-square + SER instead of a linear kernel (ablation) | Hierarchical rank pooling representation · Hollywood2 | mAP: 45.1 -> 56.8 | Helped |
Closest papers by meaning
- Generalized Rank Pooling for Activity Recognition (CVPR 2017 · similarity 0.93)
- Learning End-to-end Video Classification with Rank-Pooling (ICML 2016 · similarity 0.92)
- Non-Linear Temporal Subspace Representations for Activity Recognition (CVPR 2018 · similarity 0.91)
- Dynamic Image Networks for Action Recognition (CVPR 2016 · similarity 0.90)
- Video Representation Learning Using Discriminative Pooling (CVPR 2018 · similarity 0.88)
- HARDVS: Revisiting Human Activity Recognition with Dynamic Vision Sensors (AAAI 2024 · similarity 0.84)
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.