IS = { zkontrolovano 24 Jan 2011 },
  UPDATE  = { 2011-01-14 },
  author =     {Drbohlav, Ond{\v r}ej and Leonardis, Ale{\v s}},
  title =      {Towards correct and informative evaluation methodology 
                for texture classification under varying viewpoint 
                and illumination},
  c_title =    {Korektn{\' i} metoda pro vyhodnocen{\' i} klasifikace textur 
                nasn{\' i}man{\' y}ch z r{\accent23 u}zn{\' y}ch pohled{\accent23 u} 
                a za r{\accent23 u}zn{\' e}ho osv{\v e}tlen{\' i}},
  year =       {2010},
  month =      {April},
  pages =      {439--449},
  journal =    {Computer Vision and Image Understanding},
  publisher =  {Elsevier},
  address =    {San Diego, USA},
  issn =       {1077-3142},
  volume =     {114},
  number =     {4},
  authorship = {90-10},
  annote =     {3D texture classification under varying viewpoint and
   illumination has been a vivid research topic, and many methods have
   been developed. It is crucial that these methods be compared using
   an unbiased evaluation methodology. The most frequently employed
   methodologies use images from the Columbia-Utrecht Reflectance and
   Texture Database. These methodologies construct the training and
   test sets to be disjoint in the imaging parameters, but do not
   separate them spatially because they use images of the same surface
   patch for both. We perform a series of experiments which show that
   such practice leads to overestimation of classifier performance and
   distorts experimental findings. To correct that, we accurately
   register the images across all imaging conditions and split the
   surface patches to parts. The training and testing is then done on
   spatially disjoint parts. We show that such methodology gives a
   more realistic assessment of classifier performance. The sample
   annotations for all images are publicly available.},
  c_annote =   {Klasifikace 3D textur nasnimanych z ruznych pohledu a
   za ruzneho osvetleni je dulezitym vyzkumnym tematem. Je dulezite,
   aby metody pro klasifikaci byly vyhodnoceny pomoci metody, ktera je
   spravne zkonstruovana. Nejcasteji pouzivane metody pouzivaji
   obrazky z databaze Columbia-Utrecht Reflectance and Texure
   Database. Tyto metody pouzivaji trenovaci a testovaci data, ktera
   jsou disjunktni v parametrech popisujicich podminky snimani (smer
   pohledu, smer osvetleni), ale presto nejsou trenovaci a testovaci
   data nezavisla, protoze pouzivaji obrazky stejneho povrchu. Udelali
   jsme serii experimentu, ktera dokazuje, ze takovy pristup vede k
   vyhodnoceni, ktere je zkreslene. Abychom tento problem odstranili,
   vsechny obrazky jsme v prostorove oblasti zregistrovali a pote
   rozdelili na prostorove disjunktni casti. Metoda, ktera rozdeli
   trenovaci a testovaci data ve vsech parametrech, dava prirozene
   lepsi vyhodnoceni schopnosti testovanych algoritmu pro klasifikaci
  keywords =   {Texture classification, Illumination invariance,
                Viewpoint invariance, Evaluation methodology,
                Generalization ability},
  project =    {PERG03-GA-2008-231031 LearnTex, FP6-IST-027113, 
                MRTN-CT-2004-005439, 1M0567},