[Gzz-commits] gzz TODO

[Janne V. Kujala]

CVSROOT:        /cvsroot/gzz
Module name:    gzz
Changes by:     Janne V. Kujala <address@hidden>        03/03/03 10:00:45

Modified files:
        .              : TODO 

Log message:
        papersampler update

CVSWeb URLs:
http://savannah.gnu.org/cgi-bin/viewcvs/gzz/gzz/TODO.diff?tr1=1.604&tr2=1.605&r1=text&r2=text

Patches:
Index: gzz/TODO
diff -u gzz/TODO:1.604 gzz/TODO:1.605
--- gzz/TODO:1.604      Mon Mar  3 06:04:20 2003
+++ gzz/TODO    Mon Mar  3 10:00:44 2003
@@ -17,20 +17,18 @@
 paper paper exper:
 
     tjl:
-        - Create a script that 
-            - generates white, light, and dark versions of a unique texture 
-              for each page in http://himalia.it.jyu.fi/~jvkujala/texts.pdf
-            - saves each page with each texture type in a PNG file
-              for each of the 6 different zooms
-           - 12 textures, 12 texts, 4 zooms
-             for subject in subjects:
-               for text, texture, zoom in (texts, randomPermutation(textures),
-                                    randomPermutation(zooms + zooms + zooms)):
-                  for bg in white, light, dark:
-                     save text on texture(bg) with zoom
-            - the trials for each subject are presented in a random order
-            - the permutation of bg types for each zoom / text should
-              be evenly distributed for each subject and between subjects
+       - Fix papersampler script
+           - first image seems to incorrectly have haloed text
+           - images for second subject are broken (no text, black images, etc.)
+       - Find good set of parameters (number of texts, zooms, blend factor, 
etc.)
+         (blend factor is "blend" in the script and CONST0.a in 
pagespanpaper.py)
+        - Perhaps 18 texts, each read twice by the subject with randomly 
assigned
+         conditions?
+       - Think about randomization
+               - can we have a subject read the same text twice in a row?
+               - do we need to enforce the same number of trials for each 
ordered 
+                 condition pair corresponding to a text read twice by a 
subject?
+                 Or can we trust on number of trials to be approximately 
non-skewed?
         - Create a program that reads lines of the form
            <filename> <min-time> <max-time> <list-of-keys>
           and for each line