Research map / Architectures and efficiency / CNN design
Attention in Convolutional LSTM for Gesture 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 |
|---|---|---|---|
| remove convolutional structures from the three ConvLSTM gates (FC gates) (ablation) | Res3D+ConvLSTM+MobileNet · IsoGD RGB | top-1 accuracy: 52.01 -> 55.98 | Helped |
| remove convolutional structures from the three ConvLSTM gates (ablation) | Res3D+ConvLSTM+MobileNet · Jester | top-1 accuracy: 95.11 -> 95.12 | No clear effect |
| remove convolutional structures from gates (ablation) | Res3D+ConvLSTM+MobileNet · Parameter size | parameter size: 4.719M -> 0.529M | Helped |
| add spatial soft attention on ConvLSTM input (ablation) | Res3D+ConvLSTM+MobileNet · IsoGD RGB | top-1 accuracy: 55.98 -> 43.93 | Hurt |
| add spatial soft attention on ConvLSTM input (ablation) | Res3D+ConvLSTM+MobileNet · Jester | top-1 accuracy: 95.12 -> 94.18 | Hurt |
| replace input gate with channel-wise attention (ablation) | Res3D+ConvLSTM+MobileNet · IsoGD RGB | top-1 accuracy: 55.98 -> 53.27 | Hurt |
| replace input gate with channel-wise attention (ablation) | Res3D+ConvLSTM+MobileNet · Jester | top-1 accuracy: 95.12 -> 95.13 | No clear effect |
| replace output gate with channel-wise attention (ablation) | Res3D+ConvLSTM+MobileNet · IsoGD RGB | top-1 accuracy: 55.98 -> 54.11 | Hurt |
| replace output gate with channel-wise attention (ablation) | Res3D+ConvLSTM+MobileNet · Jester | top-1 accuracy: 95.12 -> 95.10 | No clear effect |
| use ConvLSTM variant (a) in place of ConvLSTM (ablation) | Res3D+ConvLSTM+MobileNet · IsoGD Depth | top-1 accuracy: 51.30 -> 53.28 | Helped |
| use ConvLSTM variant (a) in place of ConvLSTM (ablation) | Res3D+ConvLSTM+MobileNet · IsoGD Flow | top-1 accuracy: 45.59 -> 46.51 | Helped |
Closest papers by meaning
- Global Context-Aware Attention LSTM Networks for 3D Action Recognition (CVPR 2017 · similarity 0.89)
- LSTM Pose Machines (CVPR 2018 · similarity 0.86)
- Regularizing Long Short Term Memory With 3D Human-Skeleton Sequences for Action Recognition (CVPR 2016 · similarity 0.86)
- Regional Attention with Architecture-Rebuilt 3D Network for RGB-D Gesture Recognition (AAAI 2021 · similarity 0.86)
- An Attention Enhanced Graph Convolutional LSTM Network for Skeleton-Based Action Recognition (CVPR 2019 · similarity 0.86)
- Online Detection and Classification of Dynamic Hand Gestures With Recurrent 3D Convolutional Neural Network (CVPR 2016 · 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.