Research map / Detection and segmentation / CNN features
Object Detection As Probabilistic Set Prediction
Georg Hess, Christoffer Petersson, Lennart Svensson
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 |
|---|---|---|---|
| train with MB-NLL instead of NLL loss | RetinaNet · COCO val | PMB-NLL: 126.86 -> 121.02 | Improved |
| train with MB-NLL instead of NLL loss | Faster-RCNN · COCO val | PMB-NLL: 139.08 -> 117.77 | Improved |
| train with MB-NLL instead of NLL loss | DETR · COCO val | PMB-NLL: 152.13 -> 124.20 | Improved |
| train with MB-NLL instead of energy score | DETR · COCO val | PMB-NLL: 120.33 -> 124.20 | Hurt |
| train with MB-NLL instead of energy score | RetinaNet · COCO val | PMB-NLL: 127.66 -> 121.02 | Improved |
| train with MB-NLL instead of energy score | Faster-RCNN · COCO val | mAP: 0.373 -> 0.326 | Hurt |
| train with MB-NLL instead of NLL loss | DETR · COCO val | PDQ@F1: 0.113 -> 0.271 | Improved |
| train with MB-NLL instead of energy score | DETR · COCO val | FP term of PMB-NLL: 17.6 -> 3.1 | Improved |
| train with MB-NLL instead of energy score | DETR · COCO val | mAP: 0.407 -> 0.389 | Hurt |
| train with MB-NLL instead of NLL loss | RetinaNet · COCO val | mAP: 0.351 -> 0.361 | Improved |
| evaluate PDQ without thresholding (PDQ@0.0) instead of at F1-optimal threshold | all detectors · COCO val | PDQ: DETR ES 0.262 -> 0.033 | Hurt |
Closest papers by meaning
- Systematic Evaluation of Uncertainty Calibration in Pretrained Object Detectors (IJCV 2024 · similarity 0.89)
- COCO-O: A Benchmark for Object Detectors under Natural Distribution Shifts (ICCV 2023 · similarity 0.87)
- Estimating and Evaluating Regression Predictive Uncertainty in Deep Object Detectors (ICLR 2021 · similarity 0.87)
- Unsupervised Object Detection with Theoretical Guarantees (NeurIPS 2024 · similarity 0.87)
- Towards Evidential and Class Separable Open Set Object Detection (AAAI 2024 · similarity 0.85)
- Plug and Play Active Learning for Object Detection (CVPR 2024 · similarity 0.84)
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.
