VST-SD: Versatile Style Transfer with Content-Style Statistics Disentanglement

We introduce a self-supervised framework for (a) artistic style transfer and (b) photorealistic style transfer. We further extend it to (c) text-guided style transfer in a unified model.