Research map / Detection and segmentation / 3D detection and driving
Query2Uncertainty: Robust Uncertainty Quantification and Calibration for 3D Object Detection under Distribution Shift
Till Beemelmanns, Alexey Nekrasov, Stefan Vilceanu, Jonas Steinhaus, Timo Woopen, Bastian Leibe, Lutz Eckstein
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 density awareness to isotonic regression calibration (class-wise) | IR calibrator · nuScenes in-distribution, SECOND | D-ECE: 1.867 -> 1.839 | Improved |
| add density awareness to global isotonic regression | IR calibrator · nuScenes in-distribution, SECOND | D-ECE: 9.900 -> 5.217 | Improved |
| add density awareness to class-wise temperature scaling | TS calibrator · nuScenes in-distribution, PETR | D-ECE: 5.312 -> 5.772 | Hurt |
| add density awareness to isotonic regression under MultiCorrupt | IR calibrator · MultiCorrupt classification, PETR | D-ECE: 7.449 -> 7.211 | Improved |
| add density awareness to global isotonic regression under MultiCorrupt | IR calibrator · MultiCorrupt classification, SECOND | D-ECE: 15.689 -> 9.966 | Improved |
| add density awareness to class-wise IR on Boston to Singapore transfer | IR calibrator · Boston to Singapore, PETR | D-ECE: 5.137 -> 3.645 | Improved |
| add density awareness to class-wise TS regression under MultiCorrupt | TS calibrator · MultiCorrupt regression, SECOND | MCA_xyz: 5.264 -> 4.037 | Improved |
| add density awareness to class-wise TS regression in distribution | TS calibrator · nuScenes in-distribution regression, PETR | MCA_xyz: 1.692 -> 1.538 | Improved |
| add density awareness to global TS regression | TS calibrator · nuScenes in-distribution regression, PETR | MCA_xyz: 4.028 -> 4.059 | Hurt |
| add density awareness to TS regression on Boston to Singapore | TS calibrator · Boston to Singapore regression, PETR | MCA_xyz: 8.584 -> 8.442 | No clear effect |
| use class-wise instead of global isotonic regression (ablation) | IR calibrator · nuScenes in-distribution, PETR | D-ECE: 7.451 -> 2.999 | Improved |
| use normalizing flow instead of GMM as density estimator (ablation) | DA-IR · Distribution shift, PETR | D-ECE: 7.430 -> 7.211 | Improved |
Closest papers by meaning
- Systematic Evaluation of Uncertainty Calibration in Pretrained Object Detectors (IJCV 2024 · similarity 0.88)
- Harnessing Uncertainty-aware Bounding Boxes for Unsupervised 3D Object Detection (ICCV 2025 · similarity 0.87)
- Cocoon: Robust Multi-Modal Perception with Uncertainty-Aware Sensor Fusion (ICLR 2025 · similarity 0.87)
- Adaptive Dual Uncertainty Optimization: Boosting Monocular 3D Object Detection under Test-Time Shifts (ICCV 2025 · similarity 0.87)
- LiDAR-PTQ: Post-Training Quantization for Point Cloud 3D Object Detection (ICLR 2024 · similarity 0.84)
- Uncertainty Meets Diversity: A Comprehensive Active Learning Framework for Indoor 3D Object Detection (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.
