Research map / Label-efficient and robust learning / Self-supervised and contrastive
ImageNet-X: Understanding Model Mistakes with Factor of Variation Annotations
Read the paperCheck 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 |
|---|---|---|---|
| Increase random-crop strength | ResNet-50 · ImageNet-1k val, pose and partial view | error prevalence shift | Helped |
| Increase random-crop strength | ResNet-50 · ImageNet-1k val, subcategory | error prevalence shift | Hurt |
| Apply color-jitter augmentation | ResNet-50 · ImageNet-1k val, color and brightness | error prevalence shift | Helped |
| Apply color-jitter augmentation | ResNet-50 · ImageNet-1k val, pose | error prevalence shift | Hurt |
| Apply Gaussian-blur augmentation | ResNet-50 · ImageNet-1k val, texture and subcategory | error prevalence shift | Helped |
| Apply Gaussian-blur augmentation | ResNet-50 · ImageNet-1k val, pose | error prevalence shift | Hurt |
| Use larger training datasets | ImageNet models · ImageNet-X factors | accuracy | Helped |
| Apply robustness interventions such as CutMix, AdvProp, or AutoAugment | ImageNet models · ImageNet-X factors | per-factor accuracy | No clear effect |
| Use self-supervised rather than supervised learning | ImageNet models · ImageNet-X factors | error prevalence ratio | No clear effect |
| Use a vision transformer rather than a convolutional architecture | ImageNet models · ImageNet-X factors | error prevalence ratio | No clear effect |
Closest papers by meaning
- ImageNet-E: Benchmarking Neural Network Robustness via Attribute Editing (CVPR 2023 · similarity 0.90)
- When does dough become a bagel? Analyzing the remaining mistakes on ImageNet (NeurIPS 2022 · similarity 0.89)
- Doomed to Re-Annotate, Forever: The ImageNet Story (NeurIPS 2026 · similarity 0.88)
- Contemplating Real-World Object Classification (ICLR 2021 · similarity 0.88)
- ImageNet-Hard: The Hardest Images Remaining from a Study of the Power of Zoom and Spatial Biases in Image Classification (NeurIPS 2023 · similarity 0.87)
- ImageNet-D: Benchmarking Neural Network Robustness on Diffusion Synthetic Object (CVPR 2024 · 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.