ECE 572 Group Project Final Report: Text/image classification
for word recognition using self-organizing maps
By Arjun Shekar
1
, Chet Langin
2
and Kim Artita
3
Abstract. This paper presents a Kohonen self organizing feature map (SOM) for simple
text/image classification. The SOM is trained with 75 images containing a single word: either
“SOM”, “UTK”, or “RAM” (each word group contains 25 images with different font types).
Herein, we vary the number of weights used to train the SOM and present the resulting
prototypes. These prototypes consist of new, amalgamated/metamorphosed fonts and are
generally separated into 3 classes corresponding to the 3 word groups.
Keywords. Self Organizing Map (SOM), Self Organizing Feature Map (SOFM), Self
Organizing Map (SOM), word recognition, image classification.
1
Graduate Student, Department of Electrical & Computer Engineering, Southern Illinois University Carbondale,
Carbondale, Illinois 62901-6603
2
Graduate Student, Department of Computer Science, Southern Illinois University Carbondale, Carbondale, Illinois
62901-6603
3
Graduate Student, Department of Civil & Environmental Engineering, Southern Illinois University Carbondale,
Carbondale, Illinois 62901-6603
1