Research map / Label-efficient and robust learning / OOD and anomaly detection
Out-of-Distribution Detection with Semantic Mismatch under Masking
Yijun Yang, Ruiyuan Gao, Qiang Xu
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 MoodCat masked conditional synthesis detection compared with EBO | ResNet18 classifier · CIFAR-10 In-D, no external data | AUROC: 83.83 -> 91.39 | Improved |
| add unlabeled external ImageNet data to MoodCat training | MoodCat · CIFAR-10 In-D | FPR@TPR95: 38.83 -> 28.71 | Improved |
| add unlabeled external ImageNet data to MoodCat training | MoodCat · CIFAR-100 In-D | AUROC: 87.47 -> 89.33 | Improved |
| replace OE/UDG with MoodCat on CIFAR-100 with external data | ResNet18 classifier · CIFAR-100 In-D | FPR@TPR95: 75.45 -> 47.61 | Improved |
| use MoodCat instead of CVAECapOSR | MoodCat · CIFAR-10 6 known classes OSR | AUROC: 83.5 -> 89.48 | Improved |
| add conditional binary classifier scorer to LPIPS+DISTS (ablation) | MoodCat scoring · CIFAR-10 vs CIFAR-100 | AUROC: 74.78 -> 89.57 | Improved |
| add LPIPS and DISTS scorers to the conditional binary classifier (ablation) | MoodCat scoring · CIFAR-10 vs CIFAR-100 | AUROC: 89.13 -> 89.57 | Improved |
| add masking in training (random) versus none (ablation) | MoodCat · CIFAR-10 vs CIFAR-100 | AUROC: 90.79 -> 91.78 | Improved |
| add label conditioning to the generator (ablation) | MoodCat · CIFAR-10 In-D, CIFAR-100 OOD | AUROC: 63.62 -> 89.57 | Improved |
| add label conditioning to the generator (ablation) | MoodCat · CIFAR-100 In-D, CIFAR-10 OOD | AUROC: 52.48 -> 87.76 | Improved |
| use random masking ratio instead of fixed high ratio (ablation) | MoodCat · CIFAR-10 vs CIFAR-100 (with ImageNet) | AUROC: 91.56 -> 91.66 | Improved |
| use shuffling instead of masking (ablation) | MoodCat · CIFAR-10 vs CIFAR-100 (with ImageNet) | AUROC: 91.26 -> 88.73 | Hurt |
Closest papers by meaning
- MOOD: Multi-Level Out-of-Distribution Detection (CVPR 2021 · similarity 0.89)
- Image Background Serves as Good Proxy for Out-of-distribution Data (ICLR 2024 · similarity 0.87)
- Rethinking the Evaluation of Out-of-Distribution Detection: A Sorites Paradox (NeurIPS 2024 · similarity 0.87)
- The Best of Both Worlds: On the Dilemma of Out-of-distribution Detection (NeurIPS 2024 · similarity 0.86)
- ImageNet-OOD: Deciphering Modern Out-of-Distribution Detection Algorithms (ICLR 2024 · similarity 0.86)
- TagFog: Textual Anchor Guidance and Fake Outlier Generation for Visual Out-of-Distribution Detection (AAAI 2024 · similarity 0.85)
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.
