Research map / Architectures and efficiency / CNN design
FASON: First and Second Order Information Fusion Network for Texture Recognition
Check 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 |
|---|---|---|---|
| Add first-order mean features through the leaking shortcut (ablation) | Bilinear CNN, VGG-16 conv5 · DTD | accuracy: 72.45 -> 73.09 | Helped |
| Add first-order mean features through the leaking shortcut (ablation) | Bilinear CNN, VGG-19 conv5 · DTD | accuracy: 72.82 -> 73.62 | Helped |
| Add first-order fusion to bilinear features from conv4 and conv5 (ablation) | Bilinear CNN, VGG-16 · DTD | accuracy: 72.87 -> 74.47 | Helped |
| Add first-order fusion to bilinear features from conv4 and conv5 (ablation) | Bilinear CNN, VGG-19 · DTD | accuracy: 73.31 -> 74.57 | Helped |
| Use conv4+conv5 features instead of conv5 alone (ablation) | FASON · DTD | accuracy: 72.3 -> 72.9 | Helped |
| Use conv4+conv5 features instead of conv5 alone (ablation) | FASON · KTH-T2b | accuracy: 76.5 -> 76.4 | Hurt |
| Use conv4+conv5 features instead of conv5 alone (ablation) | FASON · MIT-Indoor, 224 input | accuracy: 76.0 -> 76.8 | Helped |
| Use conv4+conv5 features instead of conv5 alone (ablation) | FASON · Stanford Car196 | accuracy: 92.5 -> 92.8 | Helped |
| Combine first- and second-order information rather than using bilinear features alone | FASON · DTD | accuracy: 69.6 ± 0.7 -> 72.9 ± 0.7 | Helped |
| Combine first- and second-order information rather than using bilinear features alone | FASON · KTH-T2b | accuracy: 75.1 ± 2.8 -> 76.5 ± 2.3 | Helped |
| Combine first- and second-order information rather than using bilinear features alone | FASON · MIT-Indoor, 224 input | accuracy: 72.8 -> 76.8 | Helped |
| Combine first- and second-order information rather than using bilinear features alone | FASON · MIT-Indoor, 448 input | accuracy: 77.6 -> 81.7 | Helped |
Closest papers by meaning
- Two-Stream Convolutional Networks for Dynamic Texture Synthesis (CVPR 2018 · similarity 0.86)
- Deep TEN: Texture Encoding Network (CVPR 2017 · similarity 0.86)
- ImageNet-trained CNNs are not biased towards texture: Revisiting feature reliance through controlled suppression (NeurIPS 2025 · similarity 0.85)
- Harnessing Edge Information for Improved Robustness in Vision Transformers (AAAI 2024 · similarity 0.83)
- Convolutions and More as Einsum: A Tensor Network Perspective with Advances for Second-Order Methods (NeurIPS 2024 · similarity 0.83)
- Chebyshev Attention Depth Permutation Texture Network with Latent Texture Attribute Loss (CVPR 2025 · 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.