CFEW: A Large-Scale Database for Understanding Child Facial Expression in Real World
Chuangao Tang,Sunan Li,Wenming Zheng,Yuan Zong,Su Zhang,Cheng Lu,Yan Zhao
DOI: https://doi.org/10.1109/taffc.2023.3313782
IF: 13.99
2023-01-01
IEEE Transactions on Affective Computing
Abstract:Currently, much progress has been made in adult facial expressions recognition, but there has been limited attention given to the child facial expression analysis. The lack of publicly available large-scale child facial expression databases has hindered the development of automated coding for child facial expression behaviors. In this study, we aim to address this issue by constructing a new face database for understanding Child Facial Expression in real World (CFEW). The database incorporates three novelties: (1) the largest publicly available child facial expression database (11,000+ images); (2) covering the full developmental range of 0-18-year-old child subjects; (3) rich annotations for facial expression labels, including discrete categories aka happy, neutral, disgust, angry, sad, cry, fear, surprise, sleepy and other facial expressions, intensity of arousal and valence, and several types of facial action units (AUs). In addition, the images in this database capture various challenging real-world conditions, including frontal and non-frontal head poses, facial occlusions, diverse illuminations, and low image resolution. To evaluate the effectiveness of the CFEW database, we conducted extensive baseline experiments using three prominent deep convolutional neural networks: vgg11_bn, ResNet18 and DenseNet121. These experiments encompassed tasks such as discrete facial expression classification, arousal and valence estimation, facial action units detection within the database, and cross-database recognition of seven facial expressions. Both the models and the database are publicly available to the research communities, referring to http://aip.seu.edu.cn.
computer science, cybernetics, artificial intelligence