Research map / Detection and segmentation / CNN features
Mask2Anomaly: Mask Transformer for Universal Open-Set Segmentation
Shyam Nandan Rai, Fabio Cermelli, Barbara Caputo B, Carlo Masone
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 mask attention, contrastive learning and refinement mask to Mask2Former (ablation) | Mask2Former · FS L&F val | AuPRC: 10.60 -> 69.41 | Improved |
| add contrastive learning loss (to GMA + RM) (ablation) | Mask2Anomaly · FS L&F val | AuPRC: 35.05 -> 69.41 | Improved |
| add refinement mask (ablation) | Mask2Anomaly · FS L&F val | FPR95: 24.07 -> 9.46 | Improved |
| add refinement mask (ablation) | Mask2Anomaly · FS L&F val | AuPRC: 68.95 -> 69.41 | Improved |
| replace masked attention with global mask attention (ablation) | Mask2Former · FS L&F val | AuPRC: 10.60 -> 32.35 | Improved |
| replace masked attention with global mask attention (ablation) | Mask2Former · FS L&F val | mIoU: 77.42 -> 80.45 | Improved |
| replace masked attention with global mask attention (ablation) | Mask2Former · ADE20K | mIoU: 43.20 -> 43.80 | Improved |
| replace masked attention with global mask attention (ablation) | Mask2Former · Vistas | mIoU: 38.17 -> 38.92 | Improved |
| apply refinement loss on stuff excluding road instead of things excluding road (ablation) | Mask2Anomaly · FS L&F val | FPR95: 39.11 -> 9.46 | Improved |
| contrastive margin 0.75 versus 0.5 (ablation) | Mask2Anomaly · FS L&F val | AuPRC: 62.07 -> 69.41 | Improved |
| batch outlier probability 0.2 versus 0.1 (ablation) | Mask2Anomaly · FS L&F val | AuPRC: 63.01 -> 69.41 | Improved |
| add mining of unknown instances (ablation) | Mask2Anomaly · COCO open-set panoptic | PQ unknown: 9.1 -> 14.6 | Improved |
Later papers that compared with this method
2 later accepted papers report Mask2Anomaly as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Beyond Pixel Uncertainty: Bounding the OoD Objects in Road Scenes (ICCV 2025 · better on 2 of 2 shared results)
- Random Walk on Pixel Manifolds for Anomaly Segmentation of Complex Driving Scenes (ECCV 2024 · better on 3 of 3 shared results)
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
- Unmasking Anomalies in Road-Scene Segmentation (ICCV 2023 · similarity 0.94)
- Prior2Former - Evidential Modeling of Mask Transformers for Assumption-Free Open-World Panoptic Segmentation (ICCV 2025 · similarity 0.92)
- Segment Anything, Even Occluded (CVPR 2025 · similarity 0.88)
- A Dataset for Semantic Segmentation in the Presence of Unknowns (CVPR 2025 · similarity 0.87)
- PEM: Prototype-based Efficient MaskFormer for Image Segmentation (CVPR 2024 · similarity 0.86)
- Mask-Adapter: The Devil is in the Masks for Open-Vocabulary Segmentation (CVPR 2025 · 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.
