 .names file created by George John October 1994
1. TITLE:
 Image Segmentation data

2. USE IN STATLOG
 2.1 Testing Mode:
 10 Fold CrossValidation

 2.2 Special Preprocessing
 No

 2.3 Test Results
 Success Rate TIME
 Algorithm Train Test Train Test
 
 Alloc80 96.75 97.000 1248 279
 Ac2 100 96.900 3198 84
 Dipol92 ? 96.900
 BayTree ? 96.700
 NewId 100 96.600 64 69
 C4.5 98.72 96.000 142 93
 Cart 99.51 96.000 139 4
 Cn2 99.67 95.700 114 3
 IndCart 98.81 95.500 248 234
 LVQ ? 95.400
 Smart 96.13 94.800 16362 1
 BackProp ? 94.600
 Cal5 95.82 93.800 259 ?
 Kohonen ? 93.300
 Radial ? 93.100
 KNN 100 92.300 5 28
 LogDisc 90.25 89.100 302 8
 Castle 98.92 88.800 377 31
 Discrim 88.75 88.400 74 7
 QuaDisc 84.53 84.300 50 16
 Bayes 73.97 73.500 92 3
 Itrule ? 54.500
 Default ? 10.000
 Cascade ? 0.00


3. SOURCE and PASTE USAGE
 3.1 Source Information
  Creators: Vision Group, University of Massachusetts
  Donor: Vision Group (Carla Brodley, brodley@cs.umass.edu)
  Date: November, 1990

 3.2 Past Usage:
 None yet published

4. DATASET DESCRIPTION:

 The instances were drawn randomly from a database of 7 outdoor
 images. The images were handsegmented to create a classification
 for every pixel.

 Each instance is a 3x3 region.

 4.1. Number of Instances: 2310

 4.2. Number of Attributes: 19 continuous attributes

 4.3 Number of Classes: 7
 Classes:
 1 = brickface 330 examples (14.29%)
 2 = sky 330 examples (14.29%)
 3 = foliage 330 examples (14.29%)
 4 = cement 330 examples (14.29%)
 5 = window 330 examples (14.29%)
 6 = path 330 examples (14.29%)
 7 = grass 330 examples (14.29%)

 4.4. Attribute Information:

 1. regioncentroidcol: the column of the center pixel of the region.
 2. regioncentroidrow: the row of the center pixel of the region.
 3. regionpixelcount: the number of pixels in a region = 9.
 4. shortlinedensity5: the results of a line extractoin algorithm
 that counts how many lines of length 5 (any orientation) with
 low contrast, less than or equal to 5, go through the region.
 5. shortlinedensity2: same as shortlinedensity5 but counts
 lines of high contrast, greater than 5.
 6. vedgemean: measure the contrast of horizontally
 adjacent pixels in the region. There are 6, the mean and
 standard deviation are given. This attribute is used as
 a vertical edge detector.
 7. vegdesd: (see 6)
 8. hedgemean: measures the contrast of vertically adjacent
 pixels. Used for horizontal line detection.
 9. hedgesd: (see 8).
 10. intensitymean: the average over the region of (R + G + B)/3
 11. rawredmean: the average over the region of the R value.
 12. rawbluemean: the average over the region of the B value.
 13. rawgreenmean: the average over the region of the G value.
 14. exredmean: measure the excess red: (2R  (G + B))
 15. exbluemean: measure the excess blue: (2B  (G + R))
 16. exgreenmean: measure the excess green: (2G  (R + B))
 17. valuemean: 3d nonlinear transformation
 of RGB. (Algorithm can be found in Foley and VanDam, Fundamentals
 of Interactive Computer Graphics)
 18. saturationmean: (see 17)
 19. huemean: (see 17)


 Missing Attribute Values: None


NOTE
 EIGHT ATTRIBUTES ARE LINEAR COMBINATIONS (OR ARE CONSTANT), AND ARE
 BETTER REMOVED FOR THE PURPOSES OF LINEAR, QUADRATIC OR SIMILAR ALGORITHMS.
 (NAMELY attributes X3, X10X16).



CONTACTS
 statlogadm@ncc.up.pt
 bob@stams.strathclyde.ac.uk


================================================================================
1,2,3,4,5,6,7.
regioncentroidcol: continuous.
regioncentroidrow: continuous.
regionpixelcount: continuous.
shortlinedensity5: continuous.
shortlinedensity2: continuous.
vedgemean: continuous.
vegdesd: continuous.
hedgemean: continuous.
hedgesd: continuous.
intensitymean: continuous.
rawredmean: continuous.
rawbluemean: continuous.
rawgreenmean: continuous.
exredmean: continuous.
exbluemean: continuous.
exgreenmean: continuous.
valuemean: continuous.
saturationmean: continuous.
huemean: continuous.