Table 1.
EquiFACS variables (Action Units (AUs), Action Descriptors (ADs), Ear Action Descriptors (EADs) and additional variables) used in [30] after eliminating low frequency and low agreement variables.
Fig 1.
A high-level overview of the pipelines.
Fig 2.
Examples of horse faces (top) and GrayST stacked frames.
(bottom) for each of the emotional states from left to right: ‘Anticipation’, ‘Baseline’, ‘Disappointment’, ‘Frustration’.
Table 2.
Performance of the two classifier pipelines using different configurations of classes: Separation into 4 classes, 3 classes unifying Anticipation and Frustration, and just 2 classes of Anticipation and Frustration.
Table 3.
Video emotion classification results.
Fig 3.
Confusion matrix for the video-based pipeline.
Fig 4.
Confusion matrix for the Equifacs-based pipeline.
Fig 5.
The EquiFACS-based Decision Tree.