Research map / Label-efficient and robust learning / Recognition and long-tail
BiaSwap: Removing Dataset Bias With Bias-Tailored Swapping Augmentation
Eungyeup Kim, Jihyeon Lee, Jaegul Choo
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 separation of bias-contrary and bias-guiding pairs (c1) (ablation) | BiaSwap · Colored MNIST unbiased (99.0%) | accuracy: 43.04 -> 86.03 | Improved |
| add separation of bias-contrary and bias-guiding pairs (c1) (ablation) | BiaSwap · bFFHQ unbiased | accuracy: 45.0 -> 58.87 | Improved |
| add CAM-based patch sampling (c2) (ablation) | BiaSwap · Colored MNIST unbiased (99.0%) | accuracy: 84.16 -> 86.03 | Improved |
| add CAM-based patch sampling (c2) (ablation) | BiaSwap · bFFHQ unbiased | accuracy: 51.2 -> 58.87 | Improved |
| replace BiaSwap with Stylised ImageNet augmentation | vanilla classifier · Colored MNIST unbiased (99.0%) | accuracy: 48.76 -> 40.79 | Hurt |
| use BiaSwap augmentation instead of SIN | classifier · bFFHQ unbiased (99.0%) | accuracy: 69.86 -> 78.98 | Improved |
| use BiaSwap augmentation instead of vanilla training | classifier · Colored MNIST unbiased (99.0%) | accuracy: 48.76 -> 83.74 | Improved |
| use BiaSwap instead of LfF | LfF · Corrupted CIFAR10 unbiased (99.5%) | accuracy: 27.10 -> 29.11 | Improved |
| use BiaSwap instead of LfF | LfF · bFFHQ bias-guiding test | accuracy: 59.85 -> 99.13 | Improved |
| use BiaSwap instead of ReBias on real data | ReBias · BAR bias-contrary test | accuracy: 39.29 -> 52.4 | Improved |
| use BiaSwap instead of ReBias | ReBias · Colored MNIST unbiased (95.0%) | accuracy: 96.28 -> 90.85 | Hurt |
| use BiaSwap instead of vanilla | classifier · BAR bias-guiding test | accuracy: 95.00 -> 93.33 | Hurt |
Later papers that compared with this method
2 later accepted papers report BiaSwap as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Fighting Fire with Fire: Contrastive Debiasing without Bias-free Data via Generative Bias-transformation (ICML 2023 · better on 2 of 2 shared results)
- Discover and Mitigate Unknown Biases with Debiasing Alternate Networks (ECCV 2022 · 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
- BiasAdv: Bias-Adversarial Augmentation for Model Debiasing (CVPR 2023 · similarity 0.90)
- Diffusing DeBias: Synthetic Bias Amplification for Model Debiasing (NeurIPS 2025 · similarity 0.87)
- Enhancing Intrinsic Features for Debiasing via Investigating Class-Discerning Common Attributes in Bias-Contrastive Pair (CVPR 2024 · similarity 0.87)
- A Simple Remedy for Dataset Bias via Self-Influence: A Mislabeled Sample Perspective (NeurIPS 2024 · similarity 0.86)
- Classifier-to-Bias: Toward Unsupervised Automatic Bias Detection for Visual Classifiers (CVPR 2025 · similarity 0.86)
- Low-Biased General Annotated Dataset Generation (CVPR 2025 · similarity 0.86)
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.
