Research map / Detection and segmentation / Open-set and few-shot detection
Bridging Precision and Confidence: A Train-Time Loss for Calibrating Object Detection
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 BPC auxiliary loss | Deformable-DETR · COCO in-domain | D-ECE: 12.8 -> 10.3 | Helped |
| Add BPC auxiliary loss | Deformable-DETR · CorCOCO out-domain | D-ECE: 10.8 -> 9.4 | Helped |
| Add BPC auxiliary loss | Deformable-DETR · Cityscapes in-domain | D-ECE: 13.8 -> 9.9 | Helped |
| Add BPC auxiliary loss | Deformable-DETR · Foggy Cityscapes out-domain | D-ECE: 19.5 -> 12.5 | Helped |
| Add BPC auxiliary loss | Deformable-DETR · BDD100k out-domain (Cityscapes shift) | D-ECE: 11.7 -> 10.6 | Helped |
| Add BPC auxiliary loss | Deformable-DETR · Sim10k in-domain | D-ECE: 10.3 -> 6.1 | Helped |
| Set score threshold to 0.5 instead of 0.4 (ablation) | BPC · Sim10k train/test subsets | D-ECE: 9.7 -> 9.1 | Helped |
| Set score threshold to 0.6 instead of 0.5 (ablation) | BPC · Sim10k train/test subsets | D-ECE: 9.1 -> 11.2 | Hurt |
| Vary batch size from 1 to 4 (ablation) | BPC · Sim10k train/test subsets | D-ECE: 10.5; 9.1; 8.9; 10.2 | Mixed |
| Change random initialization seed (ablation) | BPC · Sim10k train/test subsets | D-ECE: 9.0; 9.1; 8.6 | No clear effect |
| Add BPC auxiliary loss | FCOS with ResNet-50 · COCO in-domain | D-ECE: 22.0 -> 20.9 | Helped |
| Add BPC auxiliary loss | FCOS with ResNet-50 · CorCOCO out-domain | D-ECE: 24.7 -> 23.5 | Helped |
Closest papers by meaning
- Multiclass Confidence and Localization Calibration for Object Detection (CVPR 2023 · similarity 0.95)
- Towards Improving Calibration in Object Detection Under Domain Shift (NeurIPS 2022 · similarity 0.93)
- Cal-DETR: Calibrated Detection Transformer (NeurIPS 2023 · similarity 0.91)
- Unsupervised Object Detection with Theoretical Guarantees (NeurIPS 2024 · similarity 0.87)
- Decoupling Feature Extraction and Classification Layers for Calibrated Neural Networks (ICML 2024 · similarity 0.86)
- A Benchmark Study on Calibration (ICLR 2024 · similarity 0.86)
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.