-> On going.

-> November 2021
-> Key words: Computational Art, Machine Learning, Pix2Pix, Chinese Calligraphy
-> Documentation

Dad, Will You Teach Me Calligraphy? is a project reflexing on the digitalization of Chinese characters, learning and machine learning, and the intimate relationship between a father and a child. This is a project that I collaborated on with my dad.

For years my dad has been self-learning calligraphy, and started to form his own style since recent years. I collected 24 pieces of his calligraphy work, trimmed them, and labeled each character artificially.

<- An example of dad’s calligraphical work.

-> Techniques

An innovative Chinese character descriping system, named Recursive Radical Packing Language (RRPL), was proposed by artist Huang Lingdong in 2018. In the system, almost all commonly-used Chinese characters can be represented by arabic numbers and symbols with a grammar system. I adopted this system to represent characters appear in the calligraphy work digitally.

The image pairs below are used as part of the inputs for the Pix2Pix conditional GAN, from which a model can be generated and exported after enough steps of training. The model is capable to transform an arbitrary character from the RRPL system to dad’s style calligraphy.

I combined the RRPL system with Markov Chain - a mathematical system that experiences transitions from one state to another according to probabilistic rules (read more here). The system hence possesses the ability to generate random symbols that look like Chinese characters but actually do not exist in Chinese dictionary, as shown below.

A “calligraphy” work was then composed with characters generated from the custom software mentioned above. The calligraphy work was passed to my dad, who rewrote it with his analogue brush and ink. The two works are presented side-by-side.