Research map / Label-efficient and robust learning / Recognition and long-tail
Deep TEN: Texture Encoding Network
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 |
|---|---|---|---|
| replace off-the-shelf FV-CNN features and encoding with end-to-end Deep-TEN | ResNet-50 on MINC-2500 · MINC-2500 | top-1 test accuracy (%): 61.8%⇒80.6% and 69.3%⇒81.3 | Helped |
| replace FV-CNN with end-to-end Deep-TEN | material and texture benchmarks · FMD, GTOS, KTH-TIPS-2b, 4D-Light, MINC-2500 | top-1 test accuracy (%) | Helped |
| replace FV-CNN with Deep-TEN | MIT-Indoor · MIT-Indoor | top-1 test accuracy (%) | Helped |
| use FV-CNN with multi-size features instead of single-size FV-CNN | FV-CNN (VGG-VD) · MINC-2500, GTOS, KTH-TIPS-2b, 4D-Light | top-1 test accuracy (%) | Helped |
| train iteratively using 352×352 and 320×320 input sizes (ablation) | Deep-TEN · MINC-2500 | top-1 test accuracy (%): 80.6 -> 81.3 | Helped |
| train iteratively using 352×352 and 320×320 input sizes (ablation) | Deep-TEN · MIT-Indoor | top-1 test accuracy (%): 71.3 -> 76.2 | Helped |
| train iteratively using 352×352 and 320×320 input sizes (ablation) | Deep-TEN · GTOS and KTH-TIPS-2b | top-1 test accuracy (%): 84.3{ $\pm$1.9} -> 84.5{ $\pm$2.9}; 82.0{ $\pm$3.3} -> 84.5{ $\pm$3.5} | Helped |
| train iteratively using 352×352 and 320×320 input sizes (ablation) | Deep-TEN · FMD and 4D-Light-Field-Material | top-1 test accuracy (%): 80.2{ $\pm 0.9$} -> 78.8{ $\pm 0.8$}; 81.7{ $\pm$1.0} -> 81.4{ $\pm$2.6} | Hurt |
| jointly train shared convolutional layers with CIFAR-10 | Deep-TEN on STL-10 · STL-10 | top-1 test accuracy (%): 76.29 -> 87.11 | Helped |
| jointly train shared convolutional layers with STL-10 | Deep-TEN on CIFAR-10 · CIFAR-10 | top-1 test accuracy (%): 91.5 -> 91.8 | Helped |
Closest papers by meaning
- Deep Structure-Revealed Network for Texture Recognition (CVPR 2020 · similarity 0.89)
- Visualizing and Understanding Deep Texture Representations (CVPR 2016 · similarity 0.88)
- Deep Texture Recognition via Exploiting Cross-Layer Statistical Self-Similarity (CVPR 2021 · similarity 0.87)
- Shape or Texture: Understanding Discriminative Features in CNNs (ICLR 2021 · similarity 0.87)
- Chebyshev Attention Depth Permutation Texture Network with Latent Texture Attribute Loss (CVPR 2025 · similarity 0.86)
- ImageNet-trained CNNs are not biased towards texture: Revisiting feature reliance through controlled suppression (NeurIPS 2025 · similarity 0.85)
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.