/[gzz]/manuscripts/Paper/paper.tex
ViewVC logotype

Diff of /manuscripts/Paper/paper.tex

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 1.27 by tjl, Mon Mar 17 19:58:10 2003 UTC revision 1.28 by jvk, Mon Mar 17 20:18:57 2003 UTC
# Line 1099  performance decreases very little over t Line 1099  performance decreases very little over t
1099    
1100  We have conducted a pilot experiment (with one subject)  We have conducted a pilot experiment (with one subject)
1101  in a similar setting comparing  in a similar setting comparing
1102  the recognition performance of our textures with pure color backgrounds.  the recognition performance of our textures with pure color backgrounds
1103  First, 15 textures (or pure colors) were shown sequentially, 5 seconds each.  (using colors from the same distribution that is used for the textures).
1104  Then, recognition was tested by showing the 15 seen textures mixed  First, 15 textures were shown sequentially, 5 seconds each.
1105    Then, recognition was tested by showing the 15 seen textures
1106  with 15 unseen textures in a random order, and the subject answered  with 15 unseen textures in a random order, and the subject answered
1107  old or new.  old or new. The test was repeated with pure color backgrounds
1108    for comparison.
1109    
1110  The results were as follows:  The results were as follows:
1111  for textures, 80 percent correct (12/15 correctly recognized  for textures, 80 percent correct (12/15 correctly recognized
1112  as seen and 12/15 correctly identified as not seen);  as seen and 12/15 correctly identified as not seen);
1113  for pure colors, 53 percent correct (13/15 correctly recognized  for pure colors, 53 percent correct (13/15 correctly recognized
1114  as seen and 3/15 correctly identified as not seen).  as seen and 3/15 correctly identified as not seen).
1115    
1116  Of course, this result is not statistically relevant,  Of course, this result is not statistically significant,
1117  but in conjunction with our experiences, it suggests that the  but in conjunction with our experiences, it suggests that the
1118  textures are quite recongizable and that they are clearly  textures are quite recongizable and that they are clearly
1119  superior to pure colors in terms of recognizability.  superior to pure colors in terms of recognizability.
1120    
1121  Furthermore, the experiment only measured the ability to  Furthermore, the experiment only measured the ability to
1122  recognize previosly seen textures within unseen textures.  recognize previosly seen textures within unseen textures.
1123  Our experience shows that in practice,  Our experience shows that
1124  at least the most recurring textures are not only recognized but can  at least the most recurring textures are not only recognized but can
1125  also easily be associated to the document content.  also easily be associated with the document content.
1126    
1127  - as per Zipf's law [XXXref], it suffices to learn the textures  - as per Zipf's law [XXXref], it suffices to learn the textures
1128    of the most recurring documents.    of the most recurring documents.
1129    
1130  XXX: refs?  XXX: refs?
1131    
1132    XXX: texture set dimensionality studies?
1133    
1134  \section{Software availability}  \section{Software availability}
1135    
1136  JVK  JVK

Legend:
Removed from v.1.27  
changed lines
  Added in v.1.28

savannah-hackers-public@gnu.org
ViewVC Help
Powered by ViewVC 1.1.26