@InProceedings{PretoFerrKura:2023:CoStAu,
               author = "Preto, Murilo de Souza and Ferreira, Fernando Teubl and Kurashima, 
                         Celso Setsuo",
          affiliation = "{Universidade Federal do ABC} and {Universidade Federal do ABC} 
                         and {Universidade Federal do ABC}",
                title = "Comparison Study of Automated Facial Expression Recognition 
                         Models",
            booktitle = "Proceedings...",
                 year = "2023",
               editor = "Clua, Esteban Walter Gonzalez and K{\"o}rting, Thales Sehn and 
                         Paulovich, Fernando Vieira and Feris, Rogerio",
         organization = "Conference on Graphics, Patterns and Images, 36. (SIBGRAPI)",
             keywords = "facial expression recognition, image processing, comparative 
                         evaluation.",
             abstract = "Facial expressions play a crucial role in human non-verbal 
                         communication, and in the psychology field there is a strong 
                         consensus on the existence of five key emotions: anger, fear, 
                         disgust, sadness, and happiness. This paper aims to evaluate 
                         multiple facial expression recognition detection models, assessing 
                         their performance across different machines and databases. By 
                         identifying the strengths and weaknesses of each option, the study 
                         seeks to comparatively determine the most suitable model for 
                         specific tasks or scenarios. For each computer, all databases were 
                         processed through the usage of the detection models, while 
                         measuring the required runtime for the facial expression 
                         detection. The detection models: Residual Masking Network and 
                         Deepface, were tested through the databases Extended Cohn-Kanade 
                         and AffectNet. The assessed data point towards an average higher 
                         accuracy for the model Residual Masking Network, but faster 
                         runtime for Deepface. Thereby, Deepface may be preferentially 
                         employed in scenarios where time constraints are a primary 
                         concern, there is limited processing capability available, or an 
                         emphasis on recognizing either happiness or neutral expressions, 
                         while Residual Masking Network might be favored in striving for a 
                         higher detection accuracy.",
  conference-location = "Rio Grande, RS",
      conference-year = "Nov. 06-09, 2023",
             language = "en",
                  ibi = "8JMKD3MGPEW34M/4B555N2",
                  url = "http://urlib.net/ibi/8JMKD3MGPEW34M/4B555N2",
           targetfile = "PretoSIBGRAPI.pdf",
        urlaccessdate = "2024, Apr. 29"
}