Research map / Label-efficient and robust learning / OOD and anomaly detection
A framework for benchmarking Class-out-of-distribution detection and its application to ImageNet
Ido Galil, Mohammed Dabbah, Ran El-Yaniv
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 knowledge distillation during training | ImageNet classifiers with matched architectures and training regimes · ImageNet C-OOD, most severity levels above s3 | AUROC | Improved |
| Pretrain on ImageNet-21k | ImageNet classifiers · ImageNet C-OOD, severity s≤3 | AUROC | Improved |
| Add adversarial training | ImageNet classifiers · ImageNet C-OOD | AUROC | No clear effect |
| Replace softmax confidence with ODIN | ImageNet classifiers · ImageNet C-OOD across severity levels | AUROC: 0.4%; 9% | Mixed |
| Replace softmax confidence with max-logit | ImageNet classifiers · ImageNet C-OOD | AUROC: +10.6% | Mixed |
| Replace softmax confidence with negative entropy | ImageNet classifiers · ImageNet C-OOD, most severity levels | AUROC | Improved |
| Replace softmax confidence with negative entropy | ImageNet classifiers · ImageNet C-OOD, severity levels s∈{7,8,9} | AUROC | Hurt |
| Add 30 dropout-enabled forward passes and predictive entropy | ImageNet classifiers · ImageNet C-OOD, especially low severity | AUROC | Improved |
| Add dropout-enabled sampling to entropy scoring (ablation) | ImageNet classifiers using MC dropout · ImageNet C-OOD, most severity levels | AUROC | Improved |
| Use CLIP cosine-similarity scores without training or fine-tuning | CLIP ResNet-50x64 · ImageNet C-OOD, severity levels below 6 | AUROC | Improved |
| Use CLIP cosine-similarity scores | CLIP ResNet-50x64 · ImageNet C-OOD, severity 10 | AUROC | Hurt |
Closest papers by meaning
- A Noisy Elephant in the Room: Is Your Out-of-Distribution Detector Robust to Label Noise? (CVPR 2024 · similarity 0.89)
- A Systematic Analysis of Out-of-Distribution Detection Under Representation and Training Paradigm Shifts (NeurIPS 2026 · similarity 0.89)
- ImageNet-OOD: Deciphering Modern Out-of-Distribution Detection Algorithms (ICLR 2024 · similarity 0.88)
- EAT: Towards Long-Tailed Out-of-Distribution Detection (AAAI 2024 · similarity 0.88)
- Image Background Serves as Good Proxy for Out-of-distribution Data (ICLR 2024 · similarity 0.87)
- Envisioning Outlier Exposure by Large Language Models for Out-of-Distribution Detection (ICML 2024 · similarity 0.87)
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.
