Research map / Video / Action recognition
Beyond Temporal Pooling: Recurrence and Temporal Convolutions for Gesture Recognition in Video
Lionel Pigou, Aäron van den Oord, Sander Dieleman, Mieke Van Herreweghe, Joni Dambre
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 |
|---|---|---|---|
| add temporal pooling to single-frame CNN | CNN · Montalbano | Jaccard index: 0.465 -> 0.775 | Improved |
| use mean-pooling instead of max-pooling over time | temporal pooling CNN · Montalbano | Jaccard index: 0.748 -> 0.775 | Improved |
| add temporal convolutions and 3D max-pooling | CNN · Montalbano | Jaccard index: 0.775 -> 0.842 | Improved |
| add bidirectional RNN on high-level spatial features | CNN · Montalbano | Jaccard index: 0.842 -> 0.888 | Improved |
| use LSTM cells instead of standard RNN cells | bidirectional RNN · Montalbano | Jaccard index: 0.885 -> 0.888 | No clear effect |
| combine temporal convolutions with LSTM RNN | Temp Conv · Montalbano | Jaccard index: 0.842 -> 0.906 | Improved |
| add skeleton-based cropping (ablation) | Temp Conv + LSTM · Montalbano | Jaccard index: 0.842 -> 0.876 | Improved |
| add depth input (ablation) | Temp Conv + LSTM (cropped) · Montalbano | Jaccard index: 0.876 -> 0.906 | Improved |
| pretrain CNN frame-by-frame before temporal max-pooling (ablation) | temporal max-pooling · Montalbano | Jaccard index | Improved |
| pretrain CNN frame-by-frame before temporal mean-pooling (ablation) | temporal mean-pooling · Montalbano | Jaccard index | No clear effect |
| use 1024 instead of 512 LSTM units (ablation) | LSTM RNN · Montalbano | validation score: 512 -> 1024 | No clear effect |
Closest papers by meaning
- Long-Term Recurrent Convolutional Networks for Visual Recognition and Description (TPAMI 2016 · similarity 0.90)
- Online Detection and Classification of Dynamic Hand Gestures With Recurrent 3D Convolutional Neural Network (CVPR 2016 · similarity 0.88)
- Temporal Convolutional Networks for Action Segmentation and Detection (CVPR 2017 · similarity 0.87)
- Memory-Augmented Temporal Dynamic Learning for Action Recognition (AAAI 2019 · similarity 0.87)
- Making Convolutional Networks Recurrent for Visual Sequence Learning (CVPR 2018 · similarity 0.87)
- Second-order Temporal Pooling for Action Recognition (IJCV 2018 · 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.
