sahtirn@ipla01.hac.com (Walter and Bill) (03/26/91)
In article <1991Mar26.171830.14240@grape.ecs.clarkson.edu> kamadoss@clutx.clarkson.edu (Shyam S. Kamadolli) writes: > > > >We are developing a neutwoor handwritten character >recognition (** aka HCR;qs**-ed)l|tmand we need a database to train and test our network. >Before spending many hours developing our own, does anybody have >or know of one we could copy? > There exists a standard database of zipcodes digitized from actual mail envelopes. It is commonly referred to as the "SUNY zipcode databese" because it is maintained by the state university of new york. In addition there is a database of digits maintained by the Arthur D. Little Co. in Washington DC. All of this boldly assumes that you are trying to solve the 10-class (digits) HCR problem. Trying to solve the more general HCR problem is probably a bit on the ambitious side. What network are you considering using?? -walter