Figure 14
From: Classification of extreme facial events in sign language videos
ESC vs. supervised classification vs. k -means results on the BU database. ESC vs. supervised classification vs. k-means clustering on the BU database for the facial events of pose yaw, pose roll, eye opening/closing and eyebrow up/down. The precision, recall and F-score percentages are averaged between the extreme states. Top: ESC performance with default configuration. Middle: ESC configuration for the best precision performance. Bottom: ESC configuration for the best recall performance.