Research map / Detection and segmentation / Open-set and few-shot detection
SAFE: Sensitivity-Aware Features for Out-of-Distribution Object Detection
Samuel Wilson, Tobias Fischer, Feras Dayoub, Dimity Miller, Niko Sünderhauf
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 |
|---|---|---|---|
| use SAFE residual+BatchNorm layers instead of a random subset of 16 other layers (ablation) | SAFE-ResNet50 · PASCAL-VOC ID, OpenImages OOD | FPR95: 66.10 -> 20.06 | Improved |
| use the 4 SAFE critical layers instead of all 60 Conv2d layers (ablation) | SAFE-ResNet50 · PASCAL-VOC ID, OpenImages OOD | FPR95: 26.73 -> 20.06 | Improved |
| use the 4 SAFE layers instead of all 12 residual connections (ablation) | SAFE-ResNet50 · PASCAL-VOC ID, OpenImages OOD | FPR95: 24.82 -> 20.06 | Improved |
| use the 4 SAFE layers instead of all 12 residual connections (ablation) | SAFE-ResNet50 · PASCAL-VOC ID, MS-COCO OOD | AUROC: 81.87 -> 80.30 | Hurt |
| train auxiliary MLP on FGSM-perturbed vs clean features instead of retraining the detector with VOS | SAFE-ResNet50 vs VOS-ResNet50 · PASCAL-VOC ID, OpenImages OOD | FPR95: 51.33 -> 20.06 | Improved |
| use SAFE instead of VOS with ResNet-50 on MS-COCO OOD (PASCAL-VOC ID) | ResNet-50 Faster-RCNN · PASCAL-VOC ID, MS-COCO OOD | AUROC: 88.70 -> 80.30 | Hurt |
| switch backbone from ResNet-50 to RegNetX4.0 for SAFE | SAFE · BDD100K ID, MS-COCO OOD | FPR95: 32.56 -> 21.69 | Improved |
| vary FGSM magnitude epsilon | SAFE-ResNet50 · PASCAL-VOC ID | AUROC and FPR95 | Mixed |
| use layers without both residual connection and BatchNorm (single random layers) (ablation) | SAFE-ResNet50 · PASCAL-VOC ID, OpenImages OOD | AUROC: 92.28 -> 75.31 | Hurt |
Closest papers by meaning
- RUNA: Object-Level Out-of-Distribution Detection via Regional Uncertainty Alignment of Multimodal Representations (AAAI 2025 · similarity 0.89)
- Leveraging Perturbation Robustness to Enhance Out-of-Distribution Detection (CVPR 2025 · similarity 0.88)
- A Systematic Analysis of Out-of-Distribution Detection Under Representation and Training Paradigm Shifts (NeurIPS 2026 · similarity 0.88)
- Image Background Serves as Good Proxy for Out-of-distribution Data (ICLR 2024 · similarity 0.88)
- A Noisy Elephant in the Room: Is Your Out-of-Distribution Detector Robust to Label Noise? (CVPR 2024 · similarity 0.87)
- Splitting & Integrating: Out-of-Distribution Detection via Adversarial Gradient Attribution (ICML 2025 · 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.
