Research map / Detection and segmentation / CNN features
SIREN: Shaping Representations for Detecting Out-of-Distribution Objects
Xuefeng Du, Gabriel Gozum, Yifei Ming, Yixuan Li
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 SIREN hyperspherical embedding shaping, KNN score | Deformable-DETR · PASCAL-VOC / MS-COCO | AUROC %: 62.15 -> 78.23 | Improved |
| use vMF score on shaped embeddings instead of Mahalanobis | Deformable-DETR · PASCAL-VOC / MS-COCO | AUROC %: 50.28 -> 76.10 | Improved |
| SIREN-KNN vs OW-DETR | Deformable-DETR · PASCAL-VOC / MS-COCO | AUROC %: 55.70 -> 78.23 | Improved |
| SIREN-KNN vs KNN | Deformable-DETR · BDD100K / MS-COCO | FPR95 %: 64.75 -> 53.97 | Improved |
| SIREN-KNN vs VOS | Faster R-CNN · PASCAL-VOC / MS-COCO | AUROC %: 88.70 -> 89.67 | Improved |
| use learnable kappa instead of fixed kappa=10 (ablation) | SIREN-KNN · PASCAL-VOC / MS-COCO | AUROC %: 77.67 -> 78.23 | Improved |
| use learned kappa instead of kappa estimated from embeddings (ablation) | SIREN-vMF · PASCAL-VOC / MS-COCO | AUROC %: 73.03 -> 76.10 | Improved |
| use nonlinear projection head instead of linear (ablation) | SIREN-vMF · PASCAL-VOC / MS-COCO | AUROC %: 71.73 -> 76.10 | Improved |
| use nonlinear projection head instead of linear (ablation) | SIREN-KNN · PASCAL-VOC / MS-COCO | AUROC %: 78.02 -> 78.23 | Improved |
| use KNN distance instead of vMF score (ablation) | SIREN · PASCAL-VOC / MS-COCO | AUROC %: 76.10 -> 78.23 | Improved |
| reduce hypersphere dimension to 8 (ablation) | SIREN-KNN · PASCAL-VOC / MS-COCO | AUROC %: 78.23 -> 77.74 | Hurt |
| increase hypersphere dimension to 80 (ablation) | SIREN-KNN · PASCAL-VOC / MS-COCO | AUROC %: 78.23 -> 76.76 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- OW-DETR — OW-DETR: Open-World Detection Transformer (CVPR 2022) · better on 2 of 2 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
- A Systematic Analysis of Out-of-Distribution Detection Under Representation and Training Paradigm Shifts (NeurIPS 2026 · similarity 0.88)
- RUNA: Object-Level Out-of-Distribution Detection via Regional Uncertainty Alignment of Multimodal Representations (AAAI 2025 · similarity 0.88)
- Is Complex Training Necessary for Long-Tailed OOD Detection? A Re-think from Feature Geometry (NeurIPS 2026 · similarity 0.87)
- Image Background Serves as Good Proxy for Out-of-distribution Data (ICLR 2024 · similarity 0.87)
- Learning to Shape In-distribution Feature Space for Out-of-distribution Detection (NeurIPS 2024 · similarity 0.87)
- YolOOD: Utilizing Object Detection Concepts for Multi-Label Out-of-Distribution Detection (CVPR 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.
