Fast Style Transfer Demo build using Google Colab notebook. This demo makes use of Tensorflow light models that I trained and converted. The tensorflow model was paired with the architecture offered by Justin Johnson et al. in their work "Perceptual Losses for Real-Time Style Transfer and Super-Resolution" as well as the Instance Normalization proposed by Dmitry Ulyanov et al. in their work "Instance Normalization: The Missing Ingredient for Fast Stylization". All training was carried out on a Google Cloud AI Platform notebook equipped with a Tesla V100 GPU and the COCO dataset.
This demo was inspired by Sayak Paul's awesome examples and his wonderful repository Adventures-in-TensorFlow-Lite, which you should definitely check out.