netcal.scaling

Scaling Methods for Confidence Calibration

This package consists of several methods for confidence calibration which use confidence scaling to approximate confidence estimates to observed accuracy. The most common scaling methods are netcal.scaling.TemperatureScaling, netcal.scaling.LogisticCalibration, and netcal.scaling.BetaCalibration. Note that all methods can also be applied to object detection and are capable of additional influenting factors such as object position and/or shape. The advanced methods netcal.scaling.LogisticCalibrationDependent and netcal.scaling.BetaCalibrationDependent are able to better represent possible correlations as the underlying probability distributions are joint multivariate distributions with possible correlations.

Available classes

AbstractLogisticRegression([method, ...])	Abstract class for all calibration methods that base on logistic regression.
LogisticCalibration(*args[, temperature_only])	On classification, apply the logistic calibration method aka Platt scaling to obtain a calibration mapping.
LogisticCalibrationDependent(*args, **kwargs)	This calibration method is for detection only and uses multivariate normal distributions to obtain a calibration mapping by means of the confidence as well as additional features.
TemperatureScaling(*args, **kwargs)	On classification or detection, apply the temperature scaling method to obtain a calibration mapping.
BetaCalibration(*args, **kwargs)	On classification, apply the beta calibration method to obtain a calibration mapping.
BetaCalibrationDependent(*args[, method])	This calibration method uses a multivariate variant of a Beta distribution to obtain a calibration mapping by means of the confidence as well as additional features.