Data for: A league-winner algorithm for defect classication in an industrial web inspection system

Data sheets with the set of defects used for training and testing the classiffiers.
Columns B to BC (if read with Excel) are the value of 54 descriptors.
Following two columns are the defect labelling given by an expert during the supervised training (both as a char or as

an index).
char E -> 6 fish-eye
char I -> 0 gel
char O -> 1 hole
char N -> 2 Black Speck
char B -> 3 Bug/Insect
char A -> 4 Wrinkle
char G -> 5 Rubber
char D -> 7 Oil Drop
char L -> 8 Light dirt
char F -> 9 Fail Transmission

First column is the name of the defect. Its first letter should be coherent with last two columns for defects with the

following exceptions:
- defects named as JXXXX.bmp are defects that can be considered as gel or rubber
- defects named as KXXXX.bmp are defects that can be considered as gel or black speck
- defects named as MXXXX.bmp are defects that can be considered as rubber or black speck