Research map / Detection and segmentation / Open-set and few-shot detection
Devil Is in the Queries: Advancing Mask Transformers for Real-World Medical Image Segmentation and Out-of-Distribution Localization
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 query-distribution loss (ablation) | MaxQuery · Pancreatic Tumors | AUROC: 63.54 -> 82.52 | Helped |
| Add query-distribution loss (ablation) | MaxQuery · Liver Tumors | AUPR: 42.31 -> 48.80 | Helped |
| Add query-distribution loss (ablation) | MaxQuery · Pancreatic Tumors val | mean inlier tumor DSC: 36.84 -> 41.77 | Helped |
| Use pre-softmax query responses rather than post-softmax query assignments (ablation) | MaxQuery · Pancreatic Tumors | AUPR: 33.98 -> 55.60 | Helped |
| Use query-level rather than category-level anomaly scores (ablation) | MaxQuery · Pancreatic Tumors | AUPR: 16.28 -> 33.98 | Helped |
| Allocate queries as (8, 4, 20) instead of (16, 4, 12) (ablation) | MaxQuery · Pancreatic Tumors | AUROC: 82.52 -> 84.44 | Helped |
| Allocate queries as (24, 4, 4) instead of (16, 4, 12) (ablation) | MaxQuery · Pancreatic Tumors | FPR95: 46.90 -> 33.70 | Helped |
| Allocate queries as (24, 4, 4) instead of (16, 4, 12) (ablation) | MaxQuery · Pancreatic Tumors | mean inlier tumor DSC: 41.77 -> 39.43 | Hurt |
Closest papers by meaning
- Interactive Medical Image Segmentation: A Benchmark Dataset and Baseline (CVPR 2025 · similarity 0.85)
- OpenMIBOOD: Open Medical Imaging Benchmarks for Out-Of-Distribution Detection (CVPR 2025 · similarity 0.84)
- S2S2: Semantic Stacking for Robust Semantic Segmentation in Medical Imaging (AAAI 2025 · similarity 0.84)
- ConDSeg: A General Medical Image Segmentation Framework via Contrast-Driven Feature Enhancement (AAAI 2025 · similarity 0.83)
- Segment Every Out-of-Distribution Object (CVPR 2024 · similarity 0.83)
- Generative Medical Segmentation (AAAI 2025 · similarity 0.83)
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.