DISFA Facial Expression Recognition Dataset

DISFA stands for Denver Intensity of Spontaneous Facial Action, which is a facial expression action dataset. The dataset consists of 27 videos, each with 4,844 frames, and a total of 130,788 images. The dataset is annotated with action units of different intensity levels.

The DISFA dataset is selected from a more popular database in the field of facial expression recognition, and contains more smiling face data (i.e., action unit 12). The smiling face action unit set in this dataset contains 30,792 images. 82,176 images in this dataset are set as action unit set, and 48,612 images are set as no action unit set.