CountSheep learns to look at an mnist digit and draw the next mnist digit counting upwards. Of course, in the process of trying to draw an image from just neural memories, CountSheep distorts its reinterpritation in sometimes chaotic and sometimes meaningful ways.
To try CountSheep for yourself and learn more about it, check out Doodle Brains.
To begin counting sheep you'll just need to install Python 3 and Pipenv which keeps track of the dependent libraries. To install them, just run
pipenv install
note: some mac's might run into this problem setting up the dependencies.
To start training the network, run
pipenv run python train.py <Number of Epochs to Train>
To watch the network generate numbers, run
pipenv run python count.py
- 0-9 reset the initial seed number to be 0 to 9
- -> Generate the next digit