Research map / Detection and segmentation / CNN features
Automated Model Evaluation for Object Detection via Prediction Consistency and Reliability
Seungju Yoo, Hyuk Kwon, Joong-Won Hwang, Kibok Lee
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 reliability score to consistency score (ablation) | PCR · Vehicle detection RetinaNet R50 corruption-based | RMSE of mAP estimate | Improved |
| add consistency score to reliability score (ablation) | PCR · Vehicle detection RetinaNet R50 corruption-based | RMSE of mAP estimate | Improved |
| add closeness of centers (CC) to IoU in the consistency score (ablation) | PCR · Vehicle detection RetinaNet R50 corruption-based | RMSE of mAP estimate | Improved |
| add confidence-based scaling that weights low-confidence predictions in consistency (ablation) | PCR · Vehicle detection RetinaNet R50 corruption-based | RMSE of mAP estimate | Improved |
| combine PCR with BoS scores in linear regression (ablation) | PCR · Vehicle detection corruption-based (4 detectors) | avg RMSE: 5.69 -> 5.15 | Improved |
| use piecewise regression with a high-mAP model on the high-mAP meta-dataset (ablation) | PCR · Vehicle detection high-mAP | avg RMSE: 5.69 -> 5.16 | Improved |
| use a low-mAP regime model on low-mAP test sets (ablation) | PCR · Vehicle detection low-mAP | avg RMSE: 13.06 -> 5.63 | Improved |
| use PCR instead of BoS on multi-class detection (ablation) | PCR · Multi-class detection corruption-based | avg RMSE: 7.96 -> 7.75 | Improved |
| use PCR instead of BoS for mAP50 (ablation) | PCR · Vehicle detection RetinaNet R50 | RMSE: 22.75 -> 10.18 | Improved |
| use PCR instead of BoS on pedestrian detection (ablation) | PCR · Pedestrian detection | avg RMSE: 6.22 -> 3.57 | Improved |
| tune hyperparameters of PCR (c=0.4, alpha=0.1, k_C=-10, k_R=20) (ablation) | PCR · RetinaNet R50 corruption-based (vehicle+pedestrian avg) | avg RMSE: 4.77 | Improved |
Closest papers by meaning
- AutoEval: Are Labels Always Necessary for Classifier Accuracy Evaluation? (TPAMI 2021 · similarity 0.88)
- Object Detection Benchmarks are Incomplete: The Role of Label Errors and Annotation Uncertainty (NeurIPS 2026 · similarity 0.86)
- Roboflow100-VL: A Multi-Domain Object Detection Benchmark for Vision-Language Models (NeurIPS 2025 · similarity 0.86)
- How to Evaluate the Generalization of Detection? A Benchmark for Comprehensive Open-Vocabulary Detection (AAAI 2024 · similarity 0.85)
- YOLOv10: Real-Time End-to-End Object Detection (NeurIPS 2024 · similarity 0.85)
- COUNTS: Benchmarking Object Detectors and Multimodal Large Language Models under Distribution Shifts (CVPR 2025 · 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.
