tom white

I'm hoping to use binarized MNIST (with [validation data](https://github.com/stitchfix/fauxtograph/pull/14)) as a sanity check to compare the NLL test score fauxtograph can achieve against other generative implementations.

I have not - my only solution so far has been to simply turn off color correction - but I would be still be interested if anyone had time to...

(note - probably needs some testing - I didn't verify it works the same as before or still works with python2)

Thanks for this, I'm glad to see someone getting some use from this library. Yes, the intention was that kerosene would handle the fuel path internally. But I think the...

Sorry for not getting to this. I still think the right fix would be to update the kerosene path is a list of strings, which I think is the newer...

I haven't written a kerosene stub for this yet, but street view house numbers format 1 is supported natively by fuel. If you would like this, I can pretty easily...

I have added a `do_mono` setting to the options which gets passed to the PixelDrawer. Currently this does two things internally: 1) RGB values are initialized to a random values...

Thanks @torridgristle - I've also copied that notebook [into this gist](https://gist.github.com/dribnet/5c222d46e3a1d9813db69b288da83311). This does look like the sort of thing I was hoping to do - perhaps also trying some noise...

I've taken pushed the `do_mono` feature a bit further, and now it outputs only 1 bit per pixel black and white.  (volcano) This is done at the last moment...

I realised that some of my thinking in my previous commit was incorrect and have redone how this is handled a bit upstream from before in a separate branch. I...