Research map / Label-efficient and robust learning / Self-supervised and contrastive
STD-Former: Image-Conditioned Texture Dictionary Encoding with Sparse Topological Supervision for Texture Recognition
Bo Peng, Ke Xu, Yurui Pan
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 naive pooling with the TDE dictionary encoder (ablation) | DINOv2 end-to-end baseline · DTD | accuracy (%): 81.2 -> 82.1 | Improved |
| add self-supervised coverage loss (single stage) (ablation) | TDE · DTD | accuracy (%): 82.1 -> 83.2 | Improved |
| use decoupled two-stage training (coverage pre-train then classifier) (ablation) | STD-Former · DTD | accuracy (%): 83.2 -> 84.5 | Improved |
| add sparse topological loss (0D/MST) in stage 2 (ablation) | STD-Former (DINOv2) · DTD | accuracy (%): 84.5 -> 86.1 | Improved |
| add sparse topological loss (0D/MST) in stage 2 (ablation) | STD-Former (DINOv2) · GTOS | accuracy (%): 90.7 -> 91.9 | Improved |
| use both intra and inter topo terms rather than one (ablation) | STD-Former+ · DTD | accuracy (%): 85.1 (intra only) / 85.0 (inter only) -> 86.1 | Improved |
| replace mean pooling of textons with attention pooling, Set Transformer PMA or concat+MLP (ablation) | STD-Former · DTD | accuracy (%): 84.5 -> 84.6 to 84.8 | No clear effect |
| switch backbone from ResNet to DINOv2 | STD-Former+ · DTD | accuracy (%): 81.7 -> 86.1 | Improved |
| add dictionary-based TDE instead of per-image KMeans prototypes | STD-Former · texture benchmarks | accuracy | Improved |
| use image-conditioned textons instead of static training-set dictionary | STD-Former · label-preserving transforms (robustness) | accuracy drop | Improved |
Closest papers by meaning
- Deep TEN: Texture Encoding Network (CVPR 2017 · similarity 0.85)
- Deep Filter Banks for Texture Recognition, Description, and Segmentation (IJCV 2016 · similarity 0.85)
- Visualizing and Understanding Deep Texture Representations (CVPR 2016 · similarity 0.84)
- Learning to Detect Features in Texture Images (CVPR 2018 · similarity 0.84)
- Chebyshev Attention Depth Permutation Texture Network with Latent Texture Attribute Loss (CVPR 2025 · similarity 0.84)
- Conditional Latent Coding with Learnable Synthesized Reference for Deep Image Compression (AAAI 2025 · similarity 0.82)
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.
