Research map / Detection and segmentation / CNN features
DeepPermNet: Visual Permutation Learning
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 |
|---|---|---|---|
| Apply Sinkhorn normalization rather than independent sigmoid outputs (ablation) | Permutation prediction model · Public Figures, sequence length 8 | Kendall Tau: 0.774 -> 0.963 | Helped |
| Apply Sinkhorn normalization rather than independent sigmoid outputs (ablation) | Permutation prediction model · Public Figures, sequence length 8 | Hamming similarity: 0.832 -> 0.973 | Helped |
| Apply Sinkhorn normalization rather than independent sigmoid outputs (ablation) | Permutation prediction model · Public Figures, sequence length 8 | Normalization error: 0.1 -> 0.022 | Helped |
| Train with longer image sequences instead of pairs (ablation) | DeepPermNet relative-attribute model · Public Figures | Mean pairwise accuracy: 7% | Helped |
| Use VGG16 instead of AlexNet (ablation) | DeepPermNet · OSR | Pairwise accuracy (Mean): 96.62 -> 98.48 | Helped |
| Use the DeepPermNet self-supervised pretext task rather than the naive permutation approach | AlexNet self-supervised pre-training · ImageNet val, 3×3 patches | Kendall Tau: 0.64 -> 0.72 | Helped |
| Use DeepPermNet rather than the naive approach for self-supervised pre-training | AlexNet transfer to PASCAL VOC 2007 · PASCAL VOC 2007 test | Classification mAP%: 65.9 -> 69.4 | Helped |
| Use DeepPermNet rather than the naive approach for self-supervised pre-training | Fast-RCNN transfer · PASCAL VOC 2007 test | FRCN Detection mAP%: 48.2 -> 49.5 | Helped |
| Use DeepPermNet rather than the naive approach for self-supervised pre-training | FCN transfer · PASCAL VOC 2012 val | FCN Segmentation mIU%: 36.8 -> 37.9 | Helped |
| Use finer 4×4 grid cells instead of the 3×3 grid (ablation) | Self-supervised DeepPermNet · Recognition tasks | Recognition performance | No clear effect |
Closest papers by meaning
- Learnable Permutation for Structured Sparsity on Transformer Models (AAAI 2026 · similarity 0.88)
- Stochastic Layer-Wise Shuffle for Improving Vision Mamba Training (ICML 2025 · similarity 0.85)
- ZipIt! Merging Models from Different Tasks without Training (ICLR 2024 · similarity 0.84)
- Towards flexible perception with visual memory (ICML 2025 · similarity 0.84)
- Beyond the Permutation Symmetry of Transformers: The Role of Rotation for Model Fusion (ICML 2025 · similarity 0.84)
- A Unified Analysis of Stochastic Gradient Descent with Arbitrary Data Permutations and Beyond (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.