Research map / Label-efficient and robust learning / OOD and anomaly detection
MOODv2: Masked Image Modeling for Out-of-Distribution Detection
Jingyao Li, Pengguang Chen, Shaozuo Yu, Shu Liu, Jiaya Jia
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 BEiT masked image modeling pretraining instead of classification ViT (ablation) | ViM score · ImageNet ID (avg 4 OOD sets) | AUROC: 77.65 -> 95.68 | Improved |
| use BEiT pretraining instead of classification ViT (ablation) | MSP score · ImageNet ID (avg 4 OOD sets) | AUROC: 76.77 -> 89.00 | Improved |
| use BEiT pretraining instead of classification ViT (ablation) | Energy score · ImageNet ID (avg 4 OOD sets) | AUROC: 65.85 -> 89.85 | Improved |
| use BEiT pretraining instead of classification ViT (ablation) | Residual score · ImageNet ID (avg 4 OOD sets) | AUROC: 74.86 -> 94.88 | Improved |
| use DINOv2 contrastive pretraining instead of classification ViT (ablation) | ViM score · ImageNet ID (avg 4 OOD sets) | AUROC: 77.65 -> 61.32 | Hurt |
| use MoCov3 contrastive pretraining instead of classification ViT (ablation) | ViM score · ImageNet ID (avg 4 OOD sets) | AUROC: 77.65 -> 72.07 | Hurt |
| use BEiT pretraining instead of classification ViT (ablation) | ViM score · CIFAR-10 ID (avg 4 OOD sets) | AUROC: 99.63 -> 99.98 | Improved |
| use BEiT pretraining instead of DINOv2 (ablation) | MSP score · CIFAR-10 ID (avg 4 OOD sets) | AUROC: 76.59 -> 59.68 | Hurt |
| MOODv2 (BEiT with ViM, no ID finetuning variant) over MOODv1 | MOOD · ImageNet ID (avg 4 OOD sets) | AUROC: 93.51 -> 95.68 | Improved |
| MOODv2 over MOODv1 | MOOD · ImageNet ID (avg 4 OOD sets) | FPR95: 31.60 -> 20.22 | Improved |
| MOODv2 over MOODv1 | MOOD · CIFAR-10 one-class (avg) | AUROC: 97.83 -> 98.20 | Improved |
| MOODv2 over MOODv1 | MOOD · CIFAR-10 one-class (avg) | FPR95: 10.33 -> 9.49 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- MOODv1 — MOOD: Multi-Level Out-of-Distribution Detection (CVPR 2021) · better on 5 of 5 shared results
- ReAct — ReAct: Out-of-distribution Detection With Rectified Activations (NeurIPS 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Image Background Serves as Good Proxy for Out-of-distribution Data (ICLR 2024 · similarity 0.90)
- A Systematic Analysis of Out-of-Distribution Detection Under Representation and Training Paradigm Shifts (NeurIPS 2026 · similarity 0.90)
- Leveraging Perturbation Robustness to Enhance Out-of-Distribution Detection (CVPR 2025 · similarity 0.88)
- Splitting & Integrating: Out-of-Distribution Detection via Adversarial Gradient Attribution (ICML 2025 · similarity 0.88)
- Diffusion-based Layer-wise Semantic Reconstruction for Unsupervised Out-of-Distribution Detection (NeurIPS 2024 · similarity 0.87)
- MultiOOD: Scaling Out-of-Distribution Detection for Multiple Modalities (NeurIPS 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.
