Rethinking Detection Calibration: A Coordinate and Direction Perspective

arXiv:2607.29040v1 Announce Type: new Abstract: Deep learning based object detectors require trustworthiness beyond competitive detection performance, but deep neural networks are prone to overconfident predictions, assigning high confidence scores to predictions that are likely to be inaccurate. To improve the alignment between confidence scores and prediction accuracy, existing methods calibrate confidence scores based on box-level localization, such as precision or intersection over union wit...