Research map / Label-efficient and robust learning / Recognition and long-tail
Addressing Model Vulnerability to Distributional Shifts Over Image Transformation Sets
Riccardo Volpi, Vittorio Murino
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 worst-case transformation search (RS) augmentation on top of random augmentation (ablation) | RDA MNIST ConvNet · MNIST RS-found transformations | accuracy: .941 -> .977 | Improved |
| add worst-case transformation search (ES) augmentation on top of random augmentation (ablation) | RDA MNIST ConvNet · MNIST ES-found transformations | accuracy: .927 -> .977 | Improved |
| add random transformation augmentation (RDA) to ERM (ablation) | ERM MNIST ConvNet · MNIST RS-found transformations | accuracy: .160 -> .941 | Improved |
| add searched augmentation (ESDA) instead of random augmentation (ablation) | RDA · MNIST to SVHN | accuracy: .395 -> .489 | Improved |
| add searched augmentation (ESDA) instead of random augmentation (ablation) | RDA · MNIST to MNIST-M | accuracy: .751 -> .816 | Improved |
| add searched augmentation (ESDA) to ERM (ablation) | ERM · MNIST to USPS | accuracy: .812 -> .840 | Improved |
| add searched augmentation (RSDA) instead of ERM (ablation) | ERM WRN-16-4 · CIFAR-10 to STL | accuracy: .466 -> .526 | Improved |
| add searched augmentation (ESDA) instead of random augmentation (ablation) | RDA WRN-16-4 · CIFAR-10 ES-found transformations | accuracy: .862 -> .913 | Improved |
| add searched augmentation (RSDA) to ERM (ablation) | ERM WRN-16-4 · CIFAR-10.1 | accuracy: .872 -> .886 | Improved |
| add searched augmentation (RSDA) instead of random augmentation (ablation) | RDA FC-DenseNet103 · CamVid RS-found transformations | pixel accuracy: .812 -> .825 | Improved |
| add robust augmentation (RDA/RSDA/ESDA) to ERM on clean data | ERM MNIST ConvNet · MNIST clean test | accuracy: .993 -> .992 | No clear effect |
| add searched augmentation (RSDA) to ERM on clean data | ERM FC-DenseNet103 · CamVid clean test | pixel accuracy: .862 -> .854 | Hurt |
Closest papers by meaning
- The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalization (ICCV 2021 · similarity 0.88)
- Measuring Robustness to Natural Distribution Shifts in Image Classification (NeurIPS 2020 · similarity 0.88)
- Evaluating Robustness to Dataset Shift via Parametric Robustness Sets (NeurIPS 2022 · similarity 0.88)
- Generalize or Detect? Towards Robust Semantic Segmentation Under Multiple Distribution Shifts (NeurIPS 2024 · similarity 0.86)
- Transformation-Invariant Learning and Theoretical Guarantees for OOD Generalization (NeurIPS 2024 · similarity 0.85)
- Adaptive Sensitivity Analysis for Robust Augmentation against Natural Corruptions in Image Segmentation (ICML 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.
