Research map / Label-efficient and robust learning / Recognition and long-tail
Examining CNN Representations With Respect to Dataset Bias
Quanshi Zhang, Wenguan Wang, Song-Chun Zhu
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 mined representation relationships rather than annotation entropy to select top-10 failure modes | AlexNet · CelebA | decrease of accuracy: 10.37 vs 25.85 | Improved |
| Use mined representation relationships rather than annotation entropy to select top-10 failure modes | AlexNet · CelebA | accuracy in ordinary images: 69.22 vs 72.29 | Mixed |
| Use mined representation relationships rather than annotation entropy to select top-40 failure modes | AlexNet · SUN Attribute | decrease of accuracy: 27.47 vs 29.68 | Improved |
| Use mined representation relationships rather than annotation entropy to select top-40 failure modes | AlexNet · SUN Attribute | accuracy in ordinary images: 63.36 vs 68.65 | Mixed |
| Reduce training-set annotation bias by adding samples with opposite attribute annotations | AlexNet trained on manually biased datasets · CelebA and SUN Attribute | KL divergence of attribute relationships | Improved |
| Use estimated representation relationships instead of annotation co-occurrence alone to identify failure modes | AlexNet · CelebA and SUN Attribute | accuracy decrease on identified failure modes | Improved |
Closest papers by meaning
- A Decade's Battle on Dataset Bias: Are We There Yet? (ICLR 2025 · similarity 0.86)
- Understanding Bias in Large-Scale Visual Datasets (NeurIPS 2024 · similarity 0.85)
- A Simple Remedy for Dataset Bias via Self-Influence: A Mislabeled Sample Perspective (NeurIPS 2024 · similarity 0.84)
- Enhancing Intrinsic Features for Debiasing via Investigating Class-Discerning Common Attributes in Bias-Contrastive Pair (CVPR 2024 · similarity 0.84)
- ImageNet-trained CNNs are not biased towards texture: Revisiting feature reliance through controlled suppression (NeurIPS 2025 · similarity 0.84)
- Classifier-to-Bias: Toward Unsupervised Automatic Bias Detection for Visual Classifiers (CVPR 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.
