Collection of tools to generate artistically interesting distortions of facial image inputs. Renders human portraits and abstract patterns in synthesized, patch-based textures. Texture synthesis algorithm is implemented from Efros & Freeman (2001).
The full pipeline allows the user to take a facial image, deconstruct it into regions either equidistant or distinguished by feature points, and render the image in a synthesized texture based on these deconstructed regions.
Individually there are also functions for facial averaging, facial warping, and texture synthesis/transfer for any source image.
| Facial Average #1 | Facial Average #2 |
|---|---|
 |
 |
| Synthesized Texture #1 | Synthesized Texture #2 |
|---|---|
 |
 |
| Transfer #1 - Larger Window Size | Transfer #2 - Smaller Window Size |
|---|---|
 |
 |
| Transfer #3 - Larger Window Size | Transfer #4 - Smaller Window Size |
|---|---|
 |
 |
- python 2.7.14
- dlib 19.1
- imageio 2.4.1
- numpy 1.15.4
- opencv 2.4.13