@InProceedings{NogueiraVeloSant:2016:StDeLe,
               author = "Nogueira, Keiller and Veloso, Adriano Alonso and Santos, Jefersson 
                         Alex dos",
          affiliation = "{Universidade Federal de Minas Gerais (UFMG)} and {Universidade 
                         Federal de Minas Gerais (UFMG)} and {Universidade Federal de Minas 
                         Gerais (UFMG)}",
                title = "Statistical and Deep Learning Algorithms for Annotating and 
                         Parsing Clothing Items in Fashion Photographs",
            booktitle = "Proceedings...",
                 year = "2016",
               editor = "Aliaga, Daniel G. and Davis, Larry S. and Farias, Ricardo C. and 
                         Fernandes, Leandro A. F. and Gibson, Stuart J. and Giraldi, Gilson 
                         A. and Gois, Jo{\~a}o Paulo and Maciel, Anderson and Menotti, 
                         David and Miranda, Paulo A. V. and Musse, Soraia and Namikawa, 
                         Laercio and Pamplona, Mauricio and Papa, Jo{\~a}o Paulo and 
                         Santos, Jefersson dos and Schwartz, William Robson and Thomaz, 
                         Carlos E.",
         organization = "Conference on Graphics, Patterns and Images, 29. (SIBGRAPI)",
            publisher = "Sociedade Brasileira de Computa{\c{c}}{\~a}o",
              address = "Porto Alegre",
             keywords = "Machine Learning, Image Annotation, Image Parsing, Descriptor, 
                         Visual Dictionary, Neural Networks, Deep Learning.",
             abstract = "Clothing identification has important roles in several areas. In 
                         this work, we present effective algorithms to automatically 
                         annotate and parse clothes from social media data. Clothing 
                         annotation tries to recognize each garment item that appears in a 
                         photo. Clothing parsing, in turn, locates and annotates each 
                         garment item in a photo. Both task pose interesting challenges for 
                         existing vision and recognition algorithms, such as distinguishing 
                         similar clothes or creating a pattern of a specific item. For the 
                         first task, two approaches, based on traditional algorithms, were 
                         proposed: (i) the pointwise one, and (ii) a multi-instance or 
                         pairwise approach. An evaluation show improvements of the proposed 
                         methods when compared to popular first choice algorithms that 
                         range from 20% to 30%. For the second task, a multi-scale 
                         convolutional network was proposed. At the end, a class is 
                         associated with each patch of the image. Experiments shows that 
                         the proposed method achieves promising results.",
  conference-location = "S{\~a}o Jos{\'e} dos Campos, SP, Brazil",
      conference-year = "4-7 Oct. 2016",
             language = "en",
                  ibi = "8JMKD3MGPAW/3M9SC2B",
                  url = "http://urlib.net/ibi/8JMKD3MGPAW/3M9SC2B",
           targetfile = "sibgrapi2016-wtd-camera_ready.pdf",
        urlaccessdate = "2024, Apr. 28"
}