Research map / Architectures and efficiency / CNN design
From BoW to CNN: Two Decades of Texture Representation for Texture Classification
Li Liu, Jie Chen, Paul Fieguth, Guoying Zhao, Rama Chellappa, Matti Pietikäinen
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 SVM instead of nearest neighbor classifier with MR8 textons | MR8 texton representation · Texture classification (2440 textons) | classification rate: 97.4% -> 98.5% | Improved |
| replace LM/S filter banks with MR8 filters | texton-based classification · texture classification | classification rate | Improved |
| replace filter bank responses with raw-patch Patch Descriptor | texton-based classification · texture classification | classification rate | Improved |
| use random projection of patches instead of raw Patch Descriptor | BoW texture classification · texture classification | classification rate: 98.5% | Improved |
| use Wavelet-based MFS instead of basic MFS | MFS · UIUC | classification rate: 92.3% -> 98.6% | Improved |
| use Wavelet-based MFS instead of basic MFS | MFS · UMD | classification rate: 93.9% -> 98.7% | Improved |
| use IFV encoder instead of BoW, LLC or VLAD on CNN features | CNN texture features · texture datasets | classification rate | Improved |
| use VGGVD+IFV instead of MRELBP | texture classification · Outex_TC10 | classification rate: 100% -> 80.0% | Hurt |
| use VGGVD+IFV instead of MRELBP | texture classification · Outex_TC12 | classification rate: 99.8% -> 82.3% | Hurt |
| pass FVCNN through a neural network to get locally transferred LFVCNN | VGGVD+IFV · DTD | classification rate: 72.3% -> 73.8% | Improved |
| pass FVCNN through a neural network to get locally transferred LFVCNN | VGGVD+IFV · KTHTIPS2b | classification rate: 81.8% -> 82.6% | Improved |
| pass FVCNN through a neural network to get locally transferred LFVCNN | VGGVD+IFV · FMD | classification rate: 79.8% -> 82.1% | Improved |
Closest papers by meaning
- Deep Filter Banks for Texture Recognition, Description, and Segmentation (IJCV 2016 · similarity 0.89)
- Visualizing and Understanding Deep Texture Representations (CVPR 2016 · similarity 0.88)
- Perceptual Texture Similarity Estimation: An Evaluation of Computational Features (TPAMI 2020 · similarity 0.88)
- Deep TEN: Texture Encoding Network (CVPR 2017 · similarity 0.88)
- Describing Textures using Natural Language (ECCV 2020 · similarity 0.87)
- ImageNet-trained CNNs are not biased towards texture: Revisiting feature reliance through controlled suppression (NeurIPS 2025 · similarity 0.84)
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.
