Research map / Detection and segmentation / Open-set and few-shot detection
Deep Filter Banks for Texture Recognition, Description, and Segmentation
Mircea Cimpoi, Subhransu Maji, Iasonas Kokkinos, Andrea Vedaldi
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 |
|---|---|---|---|
| replace SIFT with VGG-VD convolutional local descriptors | BoVW · DTD | accuracy: 52.3 -> 68.8 | Improved |
| replace SIFT with VGG-VD in Fisher vector | FV · FMD | accuracy: 59.8 -> 79.8 | Improved |
| replace FC layer with Fisher vector pooling of conv features | VGG-VD · DTD | accuracy: 62.9 -> 72.3 | Improved |
| replace FC with Fisher vector pooling of conv features | VGG-VD · MIT Indoor | accuracy: 67.6 -> 81.0 | Improved |
| replace FC with Fisher vector pooling of conv features | VGG-M · MIT Indoor | accuracy: 62.5 -> 74.2 | Improved |
| replace FC with Fisher vector pooling of conv features | VGG-VD · VOC07 | mAP: 84.6 -> 88.6 | Improved |
| use IFV or VLAD instead of BoVW or LLC pooling | VGG-M · DTD | accuracy: 61.2 -> 67.6 | Improved |
| use exp-chi2 kernel instead of linear kernel | BoVW LM · DTD | accuracy: 26.7 -> 44.0 | Improved |
| use exp-chi2 kernel instead of linear kernel | BoVW VGG-VD · DTD | accuracy: 64.1 -> 68.8 | No clear effect |
| add FC features to FV (FC+FV) | VGG-VD · FMD | accuracy: 79.8 -> 82.4 | Improved |
| add FC features to FV (FC+FV) | VGG-VD · KTH-T2b | accuracy: 81.8 -> 81.1 | No clear effect |
| replace FC with FV on CUB | VGG-VD · CUB | accuracy: 54.6 -> 66.7 | Improved |
Closest papers by meaning
- Deep TEN: Texture Encoding Network (CVPR 2017 · similarity 0.90)
- Describing Textures using Natural Language (ECCV 2020 · similarity 0.90)
- Visualizing and Understanding Deep Texture Representations (CVPR 2016 · similarity 0.90)
- Perceptual Texture Similarity Estimation: An Evaluation of Computational Features (TPAMI 2020 · similarity 0.89)
- From BoW to CNN: Two Decades of Texture Representation for Texture Classification (IJCV 2018 · similarity 0.89)
- ImageNet-trained CNNs are not biased towards texture: Revisiting feature reliance through controlled suppression (NeurIPS 2025 · 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.
