$ conda install pytorch torchvision -c soumith
Need to train one image transformation network model per one style target. According to the paper, the models are trained on the Microsoft COCO dataset.
python train.py --style_image "images/style/xyz.jpg" --dataset_path "images/content" --cuda
python fast_neural_style.py --input_image "images/input/xyz.jpg" -model "model_epoch_2.pth" --output_name "output.jpg" --cuda