Research map / Architectures and efficiency / CNN design
A Spatiotemporal Oriented Energy Network for Dynamic Texture Recognition
Isma Hadji, Richard P. Wildes
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 |
|---|---|---|---|
| use 3D spatiotemporal filtering instead of 2D (ablation) | SOE-Net · YUVL1 | accuracy: 64.0 -> 95.6 | Improved |
| use 3D spatiotemporal filtering instead of 2D (ablation) | SOE-Net · YUVL3 | accuracy: 60.8 -> 91.0 | Improved |
| use two-path rectification instead of ReLU positive path (ablation) | SOE-Net · YUVL3 | accuracy: 88.4 -> 91.0 | Improved |
| use two-path rectification instead of full wave rectification (ablation) | SOE-Net · YUVL3 | accuracy: 89.3 -> 91.0 | Improved |
| add normalization block (ablation) | SOE-Net · YUVL3 | accuracy: 82.8 -> 91.0 | Improved |
| add normalization block (ablation) | SOE-Net · YUVL1 | accuracy: 90.6 -> 95.6 | Improved |
| use Gaussian pooling instead of boxcar filter (ablation) | SOE-Net · YUVL2 | accuracy: 87.3 -> 91.7 | Improved |
| use Gaussian pooling instead of max pooling (ablation) | SOE-Net · YUVL3 | accuracy: 85.7 -> 91.0 | Improved |
| add cross-channel pooling (ablation) | SOE-Net · Dyntex_35 | accuracy: 96.8% -> 97.7% | No clear effect |
| add cross-channel pooling (ablation) | SOE-Net · Beta | accuracy: 95.1% -> 95.7% | No clear effect |
| use 3rd-order Gaussian derivative instead of 2nd (ablation) | SOE-Net · YUVL1 | accuracy: 83.3 -> 91.1 | Improved |
| use 3rd-order Gaussian derivative instead of 4th (ablation) | SOE-Net · YUVL3 | accuracy: 82.9 -> 84.6 | Improved |
Closest papers by meaning
- Dynamic Scene Recognition with Complementary Spatiotemporal Features (TPAMI 2016 · similarity 0.88)
- A New Large Scale Dynamic Texture Dataset with Application to ConvNet Understanding (ECCV 2018 · similarity 0.87)
- Temporal Residual Networks for Dynamic Scene Recognition (CVPR 2017 · similarity 0.87)
- Real-time Photorealistic Dynamic Scene Representation and Rendering with 4D Gaussian Splatting (ICLR 2024 · similarity 0.84)
- Spectral State Space Model for Rotation-Invariant Visual Representation Learning (CVPR 2025 · similarity 0.83)
- Wavelet-Driven Spatiotemporal Predictive Learning: Bridging Frequency and Time Variations (AAAI 2024 · similarity 0.83)
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.
