Research map / Label-efficient and robust learning / OOD and anomaly detection
Equipping Vision Foundation Model with Mixture of Experts for Out-of-Distribution Detection
Shizhen Zhao, Jiahui Liu, Xin Wen, Haoru Tan, Xiaojuan Qi
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 MoFE module (ablation) | DINOv2+KNN · ImageNet-1K | FPR95: 29.27 -> 22.59 | Improved |
| add Dynamic-beta Mixup (ablation) | DINOv2+KNN · ImageNet-1K | FPR95: 29.27 -> 23.85 | Improved |
| combine MoFE and Dynamic-beta Mixup (ablation) | DINOv2+KNN · ImageNet-1K | FPR95: 29.27 -> 17.01 | Improved |
| use WordNet plus K-Means refinement instead of WordNet alone for subspaces (ablation) | MoFE · ImageNet-1K | FPR95: 25.63 -> 22.59 | Improved |
| add feature-norm regularization to Mixup (ablation) | Dynamic-beta Mixup · ImageNet-1K | FPR95: 30.43 -> 23.85 | Improved |
| add dynamic beta distribution to Mixup (ablation) | Mixup with regularization · ImageNet-1K | FPR95: 24.96 -> 23.85 | Improved |
| increase number of clusters (experts) from 2 to 7 (ablation) | MoFE · ImageNet-1K | FPR95 gain: 4.10 -> 12.26 | Improved |
| increase clusters from 7 to 9 (ablation) | MoFE · ImageNet-1K | FPR95 gain: 12.26 -> 12.09 | No clear effect |
| naive fine-tune DINOv2 on ID data (ablation) | DINOv2+KNN · ImageNet-1K | FPR95: 3.01 -> 5.67 (iNaturalist); 35.30 -> 28.04 (Textures) | Mixed |
| use MoFE instead of MOS | DINOv2 · ImageNet-1K | FPR95: 26.40 -> 17.01 | Improved |
| use DINOv2+KNN instead of SigLIP features without finetuning | vision foundation models · ImageNet-1K | FPR95: 54.23 -> 29.27 | Improved |
| use MoFE with CLIP instead of NegPrompt | CLIP · ImageNet-1K | FPR95: 23.01 -> 20.02 | Improved |
Closest papers by meaning
- A Systematic Analysis of Out-of-Distribution Detection Under Representation and Training Paradigm Shifts (NeurIPS 2026 · similarity 0.90)
- Can OOD Object Detectors Learn from Foundation Models? (ECCV 2024 · similarity 0.90)
- Learning to Shape In-distribution Feature Space for Out-of-distribution Detection (NeurIPS 2024 · similarity 0.90)
- Image Background Serves as Good Proxy for Out-of-distribution Data (ICLR 2024 · similarity 0.89)
- SeTAR: Out-of-Distribution Detection with Selective Low-Rank Approximation (NeurIPS 2024 · similarity 0.88)
- Towards Optimal Feature-Shaping Methods for Out-of-Distribution Detection (ICLR 2024 · similarity 0.88)
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.
