My pytorch implementation of Style-Aware-Content-Loss-for-Real-time-HD-Style-Transfer (https://arxiv.org/abs/1807.10201), using both the original Tensorflow implementation (https://github.com/CompVis/adaptive-style-transfer) and their paper it was fairly easy to get all details needed to implement their style transfer framework.
I don't think the outputs are quite as good as the original, but it could certainly get there with a few tweak.s
More Results TBD -- its a big model with big images so takes a long time to run (>24hours on a 2080).... Add argparser etc to make more useable...