Research map / Architectures and efficiency / Attention and transformers
Attentional Pooling for 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 |
|---|---|---|---|
| Replace average pooling with attentional pooling | ResNet101 · MPII Val | mAP: 26.2 -> 30.3 | Helped |
| Add pose regularization to attentional pooling (ablation) | ResNet101 · MPII Val | mAP: 30.3 -> 30.6 | Helped |
| Replace average pooling with attentional pooling | ResNet101 · HICO | mAP: 30.2 -> 35.0 | Helped |
| Add pose regularization to attentional pooling (ablation) | ResNet101 · HICO | mAP: 35.0 -> 34.6 | Hurt |
| Replace average pooling with attentional pooling | TSN ResNet101 · HMDB51 Avg | accuracy: 47.1 -> 50.8 | Helped |
| Add pose regularization to attentional pooling (ablation) | TSN ResNet101 · HMDB51 Avg | accuracy: 50.8 -> 52.2 | Helped |
| Replace average pooling with attentional pooling | Inception-V2 · MPII Val | mAP: 25.2 -> 24.3 | Hurt |
| Use compact bilinear pooling instead of average pooling (ablation) | ResNet101 · HICO | mAP: 30.2 -> 26.8 | Hurt |
| Increase attention rank from 1 to 2 (ablation) | Attentional Pooling · MPII Val | mAP: 30.3 -> 29.9 | Hurt |
| Increase attention rank from 1 to 5 (ablation) | Attentional Pooling · MPII Val | mAP: 30.3 -> 30.0 | Hurt |
| Use class-specific bottom-up attention maps (ablation) | Attentional Pooling · MPII Val | mAP: 30.3 -> 27.9 | Hurt |
| Increase input resolution from 224px to 450px (ablation) | HMDB51 TSN ResNet · HMDB51 | accuracy | Helped |
Closest papers by meaning
- Video Representation Learning Using Discriminative Pooling (CVPR 2018 · similarity 0.88)
- AdaScan: Adaptive Scan Pooling in Deep Convolutional Neural Networks for Human Action Recognition in Videos (CVPR 2017 · similarity 0.88)
- Context-aware Attentional Pooling (CAP) for Fine-grained Visual Classification (AAAI 2021 · similarity 0.87)
- VA-AR: Learning Velocity-Aware Action Representations with Mixture of Window Attention (AAAI 2025 · similarity 0.87)
- ActionAtlas: A VideoQA Benchmark for Domain-specialized Action Recognition (NeurIPS 2024 · similarity 0.86)
- Action Detail Matters: Refining Video Recognition with Local Action Queries (CVPR 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.