Confetti Input Method / News: Recent posts

To play with the number recognition neural net:

$ make
$ ./bprop -n numbers/numbers.net -w numbers/numbers.wgt -t numbers/numbers.exm -i 1000000

Now you have a trained network with the weights stored in the file numbers.wgt

To try out some recognition:

$ ./bprop -n numbers/numbers.net -w numbers/numbers.wgt -r numbers/one.run

See the README for more information.