Oct 8, 2020
1 min read
Demonstrate AI’s Compatibility of Color/Style Changes.
See what we can do with GANs to change the photo’s color, even styles.
There are many various GAN, for different purposes and scenarios. I choose two to demo (see) how the GAN can change the photo’s color/style.
The post’s target audience is non-programmer, to help them get a GAN’s visibility/capability.
The demo is running on Colab. The code is borrowed from various open-source and modified to fit the demo’s purpose.
#1 Quick Demo for color/style change
Given a black/white photo as an input, the Demo will do:
- Determine the faces, multiple faces
- Obtained each face
- Color each face
- Cartoon style each face
- Show the colored/cartooned face
Take away points:
- The training photos will decide the color quality
- Face recognition still a challenge while the face is covered or partially showed.
- This realtime performance will also be a challenge depends on the edge/embedded devices.
- The color style can be customized by end-users.
