中国大陆用户请访问：https://www.bilibili.com/video/BV1Af4y1g7Ra/ 观看完整作品视频

The project aims to explore the combination between emergent patterns created through computer code and the potential of generating new imagery using StyleGAN – a variant of Generative Adversarial Networks (GAN). Intrigued by the topic of morphogenesis and deeply impressed by the “magic” of complex patterns emerging from a simple set of rules and formulas, I decided to explore the reaction-diffusion system and at the same time, I incorporated the Face-API to allow people to generate their reaction diffusion patterns by moving their heads around in front of a camera. Then each user can save their patterns at any time. A curated dataset of reaction diffusion system patterns is formed and created by each user in real-time. Later, this dataset is fed to the StyleGAN2-ADA model that allows for training GANs with limited data and various augmentation options. By doing so, I have an opportunity to re-look at these emergent patterns through the lens of machines and investigate how machine learning interprets these patterns and gives them new forms.

该项目旨在探索通过计算机代码创建的涌现模式与使用StyleGAN(生成对抗网络(GAN)的一种变体)生成新图像的潜力之间的结合。我被形态发生的话题所吸引，并被从一套简单的规则和公式中浮现出的复杂图案的“魔力”所深深打动，于是我决定探索反应扩散系统，同时，我加入了Face-API，让人们通过在摄像机前移动头部来生成他们的反应扩散模式。然后每个用户可以在任何时候保存他们的模式。每个用户实时生成反应扩散系统模式的精选数据集。稍后，将此数据集提供给StyleGAN2-ADA模型，该模型允许使用有限的数据和各种增强选项来训练GANs。通过这样做，我有机会通过机器的视角重新审视这些涌现的模式，并研究机器学习如何解释这些模式并赋予它们新的形式。