The paper presents the CMU-Pittsburgh AU-Coded Face Expression Image Database, a comprehensive database for facial expression analysis. It describes the problem space for facial expression analysis, which includes multiple dimensions such as level of description, transitions among expressions, eliciting conditions, reliability and validity of training and test data, individual differences in subjects, head orientation and scene complexity, image characteristics, and relation to non-verbal behavior. The database includes 2105 digitized image sequences from 182 adult subjects of varying ethnicity, performing multiple tokens of most primary FACS action units. This database is the most comprehensive test-bed to date for comparative studies of facial expression analysis. The paper discusses the challenges in facial expression analysis, including the need for fine-grained description of facial expressions, the complexity of transitions among expressions, the differences between deliberate and spontaneous expressions, the reliability of expression data, individual differences among subjects, head orientation and scene complexity, image acquisition and resolution, and the relation to non-facial behavior. It also discusses the limitations of current databases and the need for more comprehensive data that includes both deliberate and spontaneous expressions, as well as data from individuals with facial nerve damage or other impairments. The paper describes the CMU-Pittsburgh AU-Coded Face Expression Image Database, which includes 1917 image sequences from 182 subjects, with a variety of facial expressions and actions. The database has been used in several studies and provides a valuable test-bed for algorithm development and testing. The database includes a wide range of facial expressions and actions, and is designed to support the analysis of both single action units and combinations of action units. The database also includes data from infants, children, and adults, as well as data from patients with facial nerve damage or other impairments. The paper concludes that the database provides a valuable resource for the development of robust methods of facial expression analysis.The paper presents the CMU-Pittsburgh AU-Coded Face Expression Image Database, a comprehensive database for facial expression analysis. It describes the problem space for facial expression analysis, which includes multiple dimensions such as level of description, transitions among expressions, eliciting conditions, reliability and validity of training and test data, individual differences in subjects, head orientation and scene complexity, image characteristics, and relation to non-verbal behavior. The database includes 2105 digitized image sequences from 182 adult subjects of varying ethnicity, performing multiple tokens of most primary FACS action units. This database is the most comprehensive test-bed to date for comparative studies of facial expression analysis. The paper discusses the challenges in facial expression analysis, including the need for fine-grained description of facial expressions, the complexity of transitions among expressions, the differences between deliberate and spontaneous expressions, the reliability of expression data, individual differences among subjects, head orientation and scene complexity, image acquisition and resolution, and the relation to non-facial behavior. It also discusses the limitations of current databases and the need for more comprehensive data that includes both deliberate and spontaneous expressions, as well as data from individuals with facial nerve damage or other impairments. The paper describes the CMU-Pittsburgh AU-Coded Face Expression Image Database, which includes 1917 image sequences from 182 subjects, with a variety of facial expressions and actions. The database has been used in several studies and provides a valuable test-bed for algorithm development and testing. The database includes a wide range of facial expressions and actions, and is designed to support the analysis of both single action units and combinations of action units. The database also includes data from infants, children, and adults, as well as data from patients with facial nerve damage or other impairments. The paper concludes that the database provides a valuable resource for the development of robust methods of facial expression analysis.