Research map / Video / Action recognition
Discriminatively Learned Hierarchical Rank Pooling Networks
Basura Fernando, Stephen Jay Gould
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| stack recursive hierarchy on rank pooling (recursive rank pooling) (ablation) | rank pooling · Hollywood2 | mAP: 44.2 -> 52.5 | Improved |
| recursive rank pooling on HMDB51 (ablation) | rank pooling · HMDB51 | accuracy: 40.9 -> 45.8 | Improved |
| use hierarchical instead of single rank pooling (ablation) | rank pooling · Hollywood2 | mAP: 44.2 -> 56.8 | Improved |
| use SER instead of SSR feature map (ablation) | hierarchical rank pooling · Hollywood2 | mAP: 50.5 -> 56.8 | Improved |
| use SER instead of SSR feature map (ablation) | rank pooling · Hollywood2 | mAP: 44.2 -> 51.0 | Improved |
| use SER kernel instead of linear kernel (ablation) | HRP SVM · Hollywood2 | mAP: 45.1 -> 54.0 | Improved |
| combine chi-square kernel with SER (ablation) | HRP SVM · Hollywood2 | mAP: 54.0 -> 56.8 | Improved |
| use HRP instead of average pooling with chi-square+SER kernel (ablation) | pooling · Hollywood2 | mAP: 40.9 -> 56.8 | Improved |
| add HRP (CNN) to rank pooling of IDT features (ablation) | IDT RP (ALL) · Hollywood2 | mAP: 68.5 -> 74.1 | Improved |
| use HRP instead of max pooling on IDT MBH BoW (ablation) | IDT MBH · UCF101 | accuracy: 71.5 -> 82.1 | Improved |
| replace VGG-16 with ResNet features | rank pooling · UCF101 | accuracy: 72.2 -> 84.0 | Improved |
| use simple RNN or LSTM instead of max pooling (ablation) | temporal encoder · UCF101 (ResNet-152) | accuracy: 81.3 -> 75.9 | Hurt |
Closest papers by meaning
- Discriminative Hierarchical Rank Pooling for Activity Recognition (CVPR 2016 · similarity 0.97)
- Rank Pooling for Action Recognition (TPAMI 2016 · similarity 0.95)
- Learning End-to-end Video Classification with Rank-Pooling (ICML 2016 · similarity 0.94)
- Generalized Rank Pooling for Activity Recognition (CVPR 2017 · similarity 0.93)
- Action Recognition with Dynamic Image Networks (TPAMI 2017 · similarity 0.93)
- Dynamic Image Networks for Action Recognition (CVPR 2016 · similarity 0.92)
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.
