SFEW Static Facial Expression Recognition Dataset

SFEW stands for Static Facial Expression in the Wild, which is a dataset for facial expression recognition. This dataset is generated by an algorithm based on face point clustering by calculating key frames from AFEW DatabaseSelect Still Frame to create.

The most commonly used version, SFEW 2.0, is the benchmark data for the SReco sub-challenge in EmotiW 2015. SFEW 2.0 is divided into three sets: Train (958 samples), Val (436 samples), and Test (372 samples). Each image is classified into one of seven categories of expressions, namely anger, disgust, fear, neutral, happy, sad, and surprise. The labels of the training set and validation set are public, while the labels of the test set are reserved by the challenge organizers.