Just a small project to test instance segmentation with Mask R-CNN (He et al.) and neural style transfer (NST) as described in A Neural Algorithm of Artistic Style (Gatys et al.). A similar work can be found in Localized Style Transfer (Wells et al.), although I was not aware of it at the time I did this project.
Basically, first Mask R-CNN is used to detect people in an image, and then NST is applied to stylize the segmented instances. For a single instance, the algorithm is illustrated clearly in the following self-explanatory figure:
In the figure above, the artwork employed for style transfer was Vincent van Gogh's The Starry Night.
The resulting images should look something like this:
Jupyter notebook here