Table 2 Performance metrics of each 3D/2D dataset when passed into the machine learning pipeline with the 95% confidence interval provided in brackets

Training set accuracy (male)

0.776 (0.733–0.821)

0.820 (0.750–0.896)

0.891 (0.837–0.950)

0.825 (0.761–0.875)

Testing set accuracy (male)

0.636 (0.270–0.937)

0.708 (0.279–0.983)

0.704 (0.277–0.971)

0.713 (0.279–0.970)

Testing set accuracy (couple)

0.627 (0.594–0.674)

0.518 (0.406–0.594)

0.441 (0.348–0.565)

0.533 (0.442–0.630)

Testing set accuracy (female)

1.000 (1.000–1.000)

1.000 (1.000–1.000)

0.750 (0.750–0.750)

0.929 (0.750–1.000)

Testing set accuracy (focal male)

0.778 (0.583–0.833)

0.786 (0.667–0.833)

0.786 (0.667–0.833)

0.786 (0.667–0.833)

Balanced accuracy

0.656 (0.506–0.776)

0.635 (0.484–0.709)

0.588 (0.428–0.690)

0.642 (0.495–0.725)

ROC AUC

0.701 (0.618–0.763)

0.688 (0.543–0.805)

0.632 (0.452–0.774)

0.701 (0.550–0.808)

F1 (average)

0.635 (0.501–0.734)

0.597 (0.475–0.654)

0.537 (0.426–0.640)

0.604 (0.485–0.687)

F1 (male as positive class)

0.555 (0.334–0.718)

0.546 (0.317–0.672)

0.505 (0.313–0.641)

0.552 (0.323–0.695)

F1 (nonmale as positive class)

0.715 (0.662–0.756)

0.647 (0.601–0.677)

0.569 (0.525–0.642)

0.656 (0.607–0.699)

Recall (average)

0.656 (0.506–0.776)

0.635 (0.484–0.709)

0.588 (0.428–0.690)

0.642 (0.495–0.725)

Recall (male as positive class)

0.664 (0.381–0.921)

0.725 (0.386–0.961)

0.718 (0.363–0.939)

0.730 (0.386–0.95)

Recall (nonmale as positive class)

0.648 (0.616–0.692)

0.545 (0.438–0.616)

0.458 (0.370–0.575)

0.554 (0.459–0.651)

Precision (average)

0.642 (0.504–0.759)

0.629 (0.483–0.722)

0.587 (0.435–0.691)

0.635 (0.492–0.738)

Precision (male as positive class)

0.481 (0.299–0.606)

0.442 (0.271–0.538)

0.397 (0.275–0.514)

0.449 (0.281–0.568)

Precision (nonmale as positive class)

0.803 (0.665–0.931)

0.816 (0.676–0.956)

0.778 (0.595–0.937)

0.821 (0.677–0.944)