Research map / Label-efficient and robust learning / Recognition and long-tail
Stochastic Partial Swap: Enhanced Model Generalization and Interpretability for Fine-Grained Recognition
Shaoli Huang, Xinchao Wang, Dacheng Tao
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 |
|---|---|---|---|
| add Stochastic Partial Swap to mid-level branch (ablation) | M-baseline (ResNet-50) · CUB-200-2011 | accuracy (%): 82.87 -> 87.29 | Improved |
| add Stochastic Partial Swap to mid-level branch (ablation) | M-baseline (ResNet-50) · Stanford Cars | accuracy (%): 90.42 -> 94.35 | Improved |
| add Stochastic Partial Swap to mid-level branch (ablation) | M-baseline (ResNet-50) · FGVC Aircraft | accuracy (%): 89.61 -> 92.31 | Improved |
| add Stochastic Partial Swap to mid-level branch (ablation) | M-baseline (ResNet-50) · NABirds | accuracy (%): 82.85 -> 86.11 | Improved |
| add Stochastic Partial Swap to mid-level branch (ablation) | M-baseline (InceptionV3) · CUB-200-2011 | accuracy (%): 84.55 -> 86.93 | Improved |
| use SPS instead of Dropout on mid-level branch | M-baseline + Dropout (p=0.5) · CUB-200-2011 | accuracy (%): 85.48 -> 87.29 | Improved |
| use SPS instead of Cutmix on mid-level branch | M-baseline + Cutmix · CUB-200-2011 | accuracy (%): 85.71 -> 87.29 | Improved |
| sample swap ratio from U(0.3, 0.5) instead of fixed 0.4 (ablation) | SPS · CUB-200-2011 | accuracy (%): 86.65 -> 87.29 | Improved |
| add SPS to promote more regions in prediction (ablation) | M-baseline · CUB-200-2011 train set | regions contributing 99% of prediction: 35.73 -> 56.74 | Improved |
| fuse SPS mid-level branch with high-level branch | H-baseline (ResNet-50) · CUB-200-2011 | accuracy (%): 85.15 -> 88.42 | Improved |
| fuse SPS branch instead of plain mid-level baseline branch | H-baseline + M-baseline · CUB-200-2011 | accuracy (%): 87.32 -> 88.42 | Improved |
| use two SPS mid-level branches instead of one (ablation) | H-baseline + SPS · CUB-200-2011 | accuracy (%): 88.42 -> 88.70 | Improved |
Closest papers by meaning
- Fine-Grained Recognition: Accounting for Subtle Differences between Similar Classes (AAAI 2020 · similarity 0.86)
- The Unreasonable Effectiveness of Noisy Data for Fine-Grained Recognition (ECCV 2016 · similarity 0.86)
- Interpretable and Accurate Fine-grained Recognition via Region Grouping (CVPR 2020 · similarity 0.85)
- Mechanistic Permutability: Match Features Across Layers (ICLR 2025 · similarity 0.83)
- Twofold Debiasing Enhances Fine-Grained Learning with Coarse Labels (AAAI 2025 · similarity 0.83)
- Cross-Layer and Cross-Sample Feature Optimization Network for Few-Shot Fine-Grained Image Classification (AAAI 2024 · similarity 0.83)
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.
