The DET curve in assessment of detection task performance

By Alvin F. Martin, G. Doddington, T. Kamm, Mark Ordowski, Mark A. Przybocki. 1997

Abstract : We introduce the DET Curve as a means of representing performance on detection tasks that involve a tradeoff of error types. We discuss why we prefer it to the traditional ROC Curve and offer several examples of its use in speaker recognition and language recognition. We explain why it is likely to produce approximately linear curves. We also note special points that may be included on these curves, how they are used with multiple targets, and possible further applications.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Statistical Meta-Analysis of Presentation Attacks for Secure Multibiometric Systems (2016, background): “These curves report FRR vs FAR (or SFAR) for all operating points on a convenient axis scaling [15], [46].”
Deep Models and Shortwave Infrared Information to Detect Face Presentation Attacks (2020, methodology): “For this reason, results are often presented using either Receiver Operating Characteristic (ROC) or Detection-Error Tradeoff (DET) curves, which plot the APCER versus the BPCER for different thresholds [37].”
A framework for the evaluation of intrusion detection systems (2006, unknown): “An alternate approach in evaluating intrusion detection systems is to directly compare the tradeoffs in the operation of the system by a trade-off curve, such as ROC, or DET curves [21] (a reinterpretation of the ROC curve where the y -axis is 1 − P D , as opposed to P D ).”
High-level event recognition in unconstrained videos (2012, background): “[84], is often used for evaluating detection performance where the number of negative samples is significantly larger than that of the positive ones.”
Teacher-Student Training for Text-Independent Speaker Recognition (2018, methodology): “We report system performance by equal error rates (EER) and DET curves [34] based on the scores obtained from the corresponding PLDA models.”
Combining evidence from residual phase and MFCC features for speaker recognition (2006, methodology): “The performance of the speaker recognition system is given by the detection error tradeoff (DET) curve [16] shown in Fig.”
Unsupervised Domain Adaptation with a Relaxed Covariate Shift Assumption (2017, methodology): “…results based on the same metric used in Vazquez, Lopez, and Ponsa; Vazquez, Lopez, and Ponsa (2012; 2013), which is detection error tradeoff (DET) curve (Martin et al. 1997) showing the FPPI (false positive per image) rate vs. false negative (or missed detections) rate, for different thresholds.”
Learning Speaker-Specific Characteristics With a Deep Neural Architecture (2011, methodology): “For performance evaluation, we use the detection error tradeoff (DET) measure [36] to show all possible errors made in decision making during SC where the area of the operating region enclosed by a DET curve and two error axes is generally regarded as one of the best performance indexes for a detect”
A Tutorial on Text-Independent Speaker Verification (2004, background): “Furthermore, it has become a standard to plot the error curve on a normal deviate scale [47] in which case the curve is known as the detection error trade-offs (DETs) curve.”
The Geometry of ROC Space: Understanding Machine Learning Metrics through ROC Isometrics (2003, methodology): “If all models were evaluated on a test set with class ratio equal to the expected class ratio, all points in 3D ROC space are in the same horizontal plane and we use this slice as our 2D projection.”
Feature warping for robust speaker verification (2001, methodology): “The Detection Error Trade-off performance criteria [15] was selected to grade the systems examined for the evaluation.”
Relevance-Based Language Models (2001, methodology): “The tracking task in TDT is evaluated using Detection Error Tradeoff (DET) curves [13] (see figure 5).”
Improving Spoken Dialogue Understanding Using Phonetic Mixture Models (2011, background): “Discrimination between correct and incorrect responses can be viewed with a Detection Error Tradeoff (DET) curve (Martin et al. 1997).”
The BANCA Database and Evaluation Protocol (2003, methodology): “In order to visualize the performance of the system, irrespective of its operating condition, we use the conventional DET curve [9], which plots on a logdeviate scale the False Rejection Rate PFR as a function of the False Acceptance Rate PFA.”