Research map / Video / Action recognition
Heatmap Pooling Network for Action Recognition From RGB Videos
Mengyuan Liu, Jinfu Liu, Yongkang Jiang, Bin He
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 |
|---|---|---|---|
| replace 2D pose with heatmap pooled features (joint modality) | CTR-GCN · NTU RGB+D 120 X-Sub | accuracy: 83.0 -> 86.1 | Improved |
| replace 2D pose with heatmap pooled features (joint modality) | Skeleton-MixFormer · NTU RGB+D 120 X-Sub | accuracy: 82.9 -> 85.9 | Improved |
| use heatmap pooled features instead of Kinect 3D pose | CTR-GCN · NTU RGB+D 120 X-Sub | accuracy: 85.0 -> 86.1 | Improved |
| use SimCC instead of HR-Net as pose estimator (ablation) | CTR-GCN on heatmap pooled features · NTU RGB+D 120 X-Sub | accuracy: 85.3 -> 86.1 | Improved |
| pool heatmap H2 instead of H1 (ablation) | FPM · NTU RGB+D 120 X-Sub | accuracy: 81.4 -> 86.2 | Improved |
| pool heatmaps from stage 3 instead of stage 1 (ablation) | FPM with CTR-GCN · NTU RGB+D 120 X-Sub | accuracy: 69.4 -> 86.2 | Improved |
| enlarge pooling region from 1x1 to 3x3 (ablation) | FPM with CTR-GCN · NTU RGB+D 120 X-Sub | accuracy: 86.2 -> 82.0 | Hurt |
| add TRMM to FPM on Text+RGB baseline (ablation) | HP-Net · NTU RGB+D 120 X-Sub | accuracy: 95.3 -> 95.7 | Improved |
| add SMCLM on top of FPM+TRMM (ablation) | HP-Net · NTU RGB+D 120 X-Sub | accuracy: 95.7 -> 96.3 | Improved |
| use cross attention instead of TRMM to fuse pooled features (ablation) | Text+RGB baseline · NTU RGB+D 120 X-Sub | accuracy: 95.3 -> 95.4 | No clear effect |
| add human detector to filter RGB frames (ablation) | HP-Net · NTU RGB+D 120 X-Sub | accuracy: 87.0 -> 96.3 | Improved |
| vary auxiliary loss weights lambda from 0.1 to 1.0 (ablation) | HP-Net · NTU RGB+D 120 X-Sub | accuracy: 96.1 -> 96.3 | No clear effect |
Closest papers by meaning
- AdaScan: Adaptive Scan Pooling in Deep Convolutional Neural Networks for Human Action Recognition in Videos (CVPR 2017 · similarity 0.89)
- MMNet: A Model-Based Multimodal Network for Human Action Recognition in RGB-D Videos (TPAMI 2022 · similarity 0.89)
- Discriminatively Learned Hierarchical Rank Pooling Networks (IJCV 2017 · similarity 0.88)
- Discriminative Video Representation Learning Using Support Vector Classifiers (TPAMI 2019 · similarity 0.87)
- Video Representation Learning Using Discriminative Pooling (CVPR 2018 · similarity 0.87)
- NTU RGB+D: A Large Scale Dataset for 3D Human Activity Analysis (CVPR 2016 · 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.
