Game & VR 
     Race for the Arctic
     The Radiant City
     Lockdown Diary
     The Village

          Computational Creativity
     Yet It Remains Still
     JewelGAN
M
oon Feast

          Graphic Design
     Album Cover Design
Motion Graphics
         
          Performance
     DIY Orchestra
Hackoustic Iklectic

          About 

         
          Visit
     ︎ Instagram
     ︎ Vimeo
     ︎ Youtube
     ︎ Github
     ︎ LinkedIn

JewelGAN


A Jewellery GAN Model / Customised Ring Dataset / A experimental website design(WIP)





Crafting a metal ring takes patience, ingenuity and a master-hand. Time and fortune is indispensable if you want to turn a metal ring into a unique relic, waiting for it get through lengthy history and being polished by time. However, recent advancement in machine learning techniques offer a possible accelerating solution. Generative Adversarial Networks(GANs) and Neural Style Transfer could help transform existing jewellery items into relic items by its extraordinary capabilities in image generation. I proposed JewelGan: an Wasserstein generative adversarial network application, which can generate inexistent relic jewelleries conditioned on specific textures. The result of JewelGAN offer an initiate glance of how AI image generation algorithms and computational creativity philosophy can be used to the field of culture heritage and archaeology researches.

The source code  can be found at: https://github.com/feiqiwangbot/JewelGAN.







Screenshots of datasets

As an important component of archaeological research, metalware have an extremely long life cycle comparing with other biodegradable material. These metallic items especially jewellery pieces which were carefully kept by its owner, have always survived from lengthy history evolution and been unearthed by people in their future. When displayed in museum historical collections, these pieces show their extraordinary beauty that has been polished by time.

In this project, I try to introduce generative adversarial network(GANs) and generative deep learning technique of neural style transfer into the field of archaeology, and consider these machine learning techniques as jewellery craftsmen and a time accelerator, that help me transform modern jewellery items into rusty historical relic which only possibly happens in a very distant future. The project attempts to create a fictional context for future archaeology. Ideally, the final outcome of this project is to build a website for a virtual jewellery gallery. Audiences can visit the website and  browse these artificial relic jewellery images like visiting an online museum. There will also be a preorder link for audiences to reserve a relic piece, though it will only be available for delivery in a 20000+ years future. By building this virtual gallery of non-existent jewellery, I try to discuss the value of precious objects like rings as they have been stripped of their ornate appearance over time but given a historical quality, and I want to experiment how artificial intelligence technique and computational creativity could engage in this creative process. 



Unlimited rings to come..