@InProceedings{SilvaNasc:2017:ReInSc,
               author = "Silva, Camila Laranjeira da and Nascimento, Erickson Rangel",
          affiliation = "{Universidade Federal de Minas Gerais} and {Universidade Federal 
                         de Minas Gerais}",
                title = "Representing Indoor Scenes as a Sparse Composition of Feature 
                         Segments",
            booktitle = "Proceedings...",
                 year = "2017",
               editor = "Torchelsen, Rafael Piccin and Nascimento, Erickson Rangel do and 
                         Panozzo, Daniele and Liu, Zicheng and Farias, Myl{\`e}ne and 
                         Viera, Thales and Sacht, Leonardo and Ferreira, Nivan and Comba, 
                         Jo{\~a}o Luiz Dihl and Hirata, Nina and Schiavon Porto, Marcelo 
                         and Vital, Creto and Pagot, Christian Azambuja and Petronetto, 
                         Fabiano and Clua, Esteban and Cardeal, Fl{\'a}vio",
         organization = "Conference on Graphics, Patterns and Images, 30. (SIBGRAPI)",
            publisher = "Sociedade Brasileira de Computa{\c{c}}{\~a}o",
              address = "Porto Alegre",
             keywords = "Indoor Scene Recognition, Semantic Segmentation, Regularization.",
             abstract = "Researchers in the fields of Computer Vision and Pattern 
                         Recognition have been trying to tackle the problem of scene 
                         recognition for many years. Several approaches rely on the 
                         assumption that object-level information can be highly 
                         discriminatory, which has been extensively validated in the 
                         literature. We propose an approach that merges sparse semantic 
                         segmentation features with object features, composing a sparse 
                         representation of feature segments, as an attempt to represent the 
                         composition of objects of a given scene. Our premise is that by 
                         adding sparsity constraints to a semantic segmentation feature, we 
                         represent a small amount of well chosen objects or parts of 
                         objects. We expect this will add robustness to the final feature, 
                         since it will recognize a given scene by its most distinctive 
                         segments, thus increasing the generalization power of the 
                         representation. According to our results, the methodology seems 
                         promising, but it is strongly affected by the poor performance of 
                         segmentation features on classes containing small objects.",
  conference-location = "Niter{\'o}i, RJ, Brazil",
      conference-year = "17-20 Oct. 2017",
             language = "en",
                  ibi = "8JMKD3MGPAW/3PJ55BB",
                  url = "http://urlib.net/ibi/8JMKD3MGPAW/3PJ55BB",
           targetfile = "Sibgrapi_2017_WiP_camera-ready.pdf",
        urlaccessdate = "2024, May 01"
}