Machine Learning x Animation: Yuru Kyara
Project details
- Year
- 2022
- Programme
- animation
- Practices
- Autonomous
- Minor
- Hacking
Project Yuru Kyara was born out of my desire to be less rigid and more playful in the artistic process, through exploring the use of machine learning as a collaborative ‘partner’, known for its unpredictability with the focus being my interest in (hand drawn) character animation.
It will be partly incorporated in to some shots of our graduation film: ZEZE & MERCEDES.
“Yuru-chara (Japanese: ゆるキャラ, Hepburn: yuru kyara) is a Japanese term for a category of mascot characters; usually created to promote a place or region, event, organisation or business. They are characterized by their kawaii (cute) and unsophisticated designs, often incorporating motifs that represent local culture, history or produce.”
The initial concept for our graduation film ZEZE & MERCEDES at the time provided the fertile ground, which inspired my datasets of Japanese mascots, the would-be imaginary creatures inhabiting the parade of Zeze and Mercedes’ fantasy play.
These datasets were then used to train two machine learning models: StyleGAN and Pix2PixHD. The surrealistic quality of the generated mascot characters by the StyleGAN combined with the hand drawn component introduced by Pix2PixHD formed the basis of this project.
Special thanks to Boris Smeenk for the technical support.
To learn more about my machine learning explorations and datasets, visit yurukyarazm.weebly.com
