Research map / Label-efficient and robust learning / Recognition and long-tail
SnapMix: Semantically Proportional Mixing for Augmenting Fine-grained Data
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 |
|---|---|---|---|
| Use semantic-ratio label weights instead of area-ratio label weights (ablation) | SnapMix image mixing on fine-grained recognition · CUB, Cars, and Aircraft | top-1 accuracy | Improved |
| Use asymmetric rather than symmetric image mixing (ablation) | SnapMix · CUB, Cars, and Aircraft | top-1 accuracy | Improved |
| Set alpha to 5.0 rather than other tested values (ablation) | SnapMix on ResNet-50 · CUB | accuracy: 87.22, 87.23, 87.25, 87.30, 87.75, 87.30, 87.54 | Improved |
| Apply SnapMix rather than the strongest listed mixing augmentation | ResNet-18 and ResNet-34 · CUB, Cars, and Aircraft | top-1 accuracy | Improved |
| Apply SnapMix rather than the strongest listed mixing augmentation | ResNet-50 and ResNet-101 · CUB, Cars, and Aircraft | top-1 accuracy | Improved |
| Apply CutMix rather than the baseline | ResNet-18 · CUB | top-1 accuracy: 82.35 -> 80.16 | Hurt |
| Apply CutOut rather than the baseline | ResNet-18 · CUB | top-1 accuracy: 82.35 -> 80.54 | Hurt |
| Apply SnapMix rather than the baseline | ResNet-18 · CUB, training from scratch | accuracy: 64.98 -> 70.31 | Improved |
| Apply SnapMix rather than MixUp | ResNet-50 · CUB, training from scratch | accuracy: 72.39 -> 72.17 | Hurt |
| Apply SnapMix rather than the baseline | InceptionV3 and DenseNet121 · CUB | accuracy | Improved |
Later papers that compared with this method
1 later accepted paper reports SnapMix as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Explanation-based Data Augmentation for Image Classification (NeurIPS 2021 · better on 2 of 2 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- DiffuseMix: Label-Preserving Data Augmentation with Diffusion Models (CVPR 2024 · similarity 0.86)
- TokenMix: Rethinking Image Mixing for Data Augmentation in Vision Transformers (ECCV 2022 · similarity 0.86)
- TransMix: Attend To Mix for Vision Transformers (CVPR 2022 · similarity 0.86)
- Adversarial AutoMixup (ICLR 2024 · similarity 0.85)
- RegMixMatch: Optimizing Mixup Utilization in Semi-Supervised Learning (AAAI 2025 · similarity 0.84)
- Enhance Image Classification via Inter-Class Image Mixup with Diffusion Model (CVPR 2024 · 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.
