Local Image to Image Translation Via Pixel wise Highway Adaptive Instance Normalization

Demo

• A demo page is in the following link : http://123.108.168.4:5000/

Usage

Prerequisite

• python3.6
• Pytorch0.4
• matplotlib 2.2.2

To download the CelebA:

$ bash data_download.sh

Train

• We have used a jupyter notebook for the training. (An example is in "Hair color, Non-smile.ipynb")

The result of LOcal Mask based Image Translation (LOMIT)

Comparison with StarGAN (black, smile) => (blonde, non-smile)

From the top, each row denotes input image, an exemplar and a corresponding output.

Other results (Facial Hair and Gender translation)

First row of each macro row denotes an input image and the second row indicates an exemplar. The third row is a corresponding output.

---

---