Research map / Detection and segmentation / Open-set and few-shot detection
GL-MCM: Global and Local Maximum Concept Matching for Zero-Shot Out-of-Distribution Detection
Atsuyuki Miyai, Qing Yu, Go Irie, Kiyoharu Aizawa
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 global score to local score (GL-MCM instead of L-MCM alone) (ablation) | L-MCM · ImageNet-1K (avg) | FPR95: 73.59 -> 35.25 | Improved |
| add local L-MCM score to MCM (ablation) | MCM · ImageNet-1K (avg) | FPR95: 42.74 -> 35.25 | Improved |
| add local L-MCM score to MCM (ablation) | MCM · MS-COCO (avg) | FPR95: 73.96 -> 62.21 | Improved |
| add local L-MCM score to MCM (ablation) | MCM · Pascal-VOC (avg) | FPR95: 38.88 -> 31.12 | Improved |
| add local score to MCM on Texture OOD set (ViT-B/16) | MCM · ImageNet Texture | AUROC: 86.11 -> 85.54 | Hurt |
| softmax temperature 1.0 instead of 0.01 (ablation) | GL-MCM · ImageNet-1K (avg) | FPR95: 50.72 -> 35.25 | Improved |
| softmax temperature 1.0 instead of 0.01 (ablation) | GL-MCM · MS-COCO (avg) | FPR95: 64.89 -> 62.21 | Improved |
| apply GL-MCM score to prompt-learned CoOp | CoOp · ImageNet-1K (avg) | FPR95: 36.83 -> 30.03 | Improved |
| apply GL-MCM score to prompt-learned LoCoOp | LoCoOp · ImageNet-1K (avg) | FPR95: 33.98 -> 28.68 | Improved |
| use CLIP ResNet-50 backbone with GL-MCM instead of MCM | MCM · ImageNet-1K (avg, ResNet-50) | FPR95: 49.66 -> 40.86 | Improved |
| add global score to L-MCM when local features come from SAN (ablation) | L-MCM with SAN · ImageNet-1K | FPR95: 48.29 -> 32.91 | Improved |
| replace sum of MCM and L-MCM by max of the two scores (ablation) | GL-MCM · ImageNet-1K | FPR95: 35.25 -> 51.5 | Hurt |
Closest papers by meaning
- Intra-Image Mining and Symmetric Maximum Concept Matching for Few Shot Out-of-Distribution Detection (AAAI 2026 · similarity 0.90)
- Delving into Out-of-Distribution Detection with Vision-Language Representations (NeurIPS 2022 · similarity 0.89)
- Test-Time Adaptation via Self-Reinforced Optimal Transport for Zero-Shot OOD Detection with Vision–Language Models (NeurIPS 2026 · similarity 0.87)
- How Does Fine-Tuning Impact Out-of-Distribution Detection for Vision-Language Models? (IJCV 2023 · similarity 0.87)
- Conjugated Semantic Pool Improves OOD Detection with Pre-trained Vision-Language Models (NeurIPS 2024 · similarity 0.87)
- GIM: Learning Generalizable Image Matcher From Internet Videos (ICLR 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.
