Neural Style Transfer

This notebook contains a typical machine learning/computer vision problem: Neural Style Transfer. With a content image (C) and a style image (S), we are able to generate a image (G) with C’s content and S’s style. The following images show how a photo can be combined with Van Gogh’s painting and generate a new image.

drawing

Content Image: C

drawing

Style Image: S

drawing

Generated Image: G

Acknowledgement:

Some ideas come from Deep Learning Course on Coursera (e.g., the weights of style loss and content loss) and Pytorch tutorial (e.g. model loading).