Journal: PLOS ONE
Article Title: Training machine learning algorithms for automatic facial coding: The role of emotional facial expressions’ prototypicality
doi: 10.1371/journal.pone.0281309
Figure Lengend Snippet: Note . Action Unit (AU) activity in both datasets was extracted with Noldus FaceReader (FR, Version 7.1). The training process for both datasets was identical: preprocessing, split into 70% training dataset and 30% test dataset, hyperparameter optimization with grouped 10-fold cross-validation in the training dataset. Models trained on each dataset were evaluated twice: once with the 30% hold-out test dataset from the corresponding training dataset and once with 100% of the respective other dataset. The picture for the standardized dataset is model AF33HAS from the KDEF database . The exemplary picture for the unstandardized dataset symbolizes but is not taken from one of our anonymous participants; the model provided written informed consent.
Article Snippet: Both datasets were processed with FaceReader (FR, Version 7.1, Noldus Information Technology) [ ].
Techniques: Activity Assay, Biomarker Discovery