This is a small selection of the images, videos, and even 3d models created by students during the course. We had an AMAZING group of people. Even more important than the creation of digital artifacts were the lively discussions about our fears and hopes for machine learning and artificial intelligence, issues in ML/AI such as job displacement… Continue reading Machine Learning for Artists course wrap-up
50 million doodles have been released as a dataset, including 126,475 drawings of a “vase”. In this image I’ve stacked hundreds of “vase” doodles, revealing a common form. But what emerges is not the universal, Platonic ideal of a “vase”. Google’s analysis of doodle composites has shown that even for basic things like “chair” there are significant… Continue reading What is a vase?
The virtual is slowly seeping into the real. @pointyview and @gingerlukas have been posting some wonderful tests of glaze recipes generated by OpenAI’s machine learning software GPT-2. This is only the beginning!
Class full! Thanks so much everyone. Really excited to see what people create.
A selection of machine-learning generated Dehua porcelain heads. StyleGAN2-ada-pytorch on Google Colab with single NVIDIA V100 GPU. Converted legacy network FFHQ 512×512 StyleGAN2 config f, then using 500 images of faces from dehua sculptures ran transfer learning for 200kimg (default augmentation, x-mirror).
These quotes were genearated by a huge language model from OpenAI called GPT-3. After prompting the model with real quotes by famous artists like “Art is a line around your thoughts” (Gustav Klimt) and “To be an artist is to believe in life” (Henry Moore), GPT-3 responded with a wide range of its own quotes. A small… Continue reading This inspirational art quote does not exist.
Colorize black & white photos with machine learning for free! Recently, the website My Heritage has made their automatic photo colorization service available for free (although you must still sign up for a free account). They are using machine learning algorithms trained against millions of photos, and the results are pretty amazing. EDIT: Just looked into… Continue reading Colorize!
Thanks so much @ceramicreview for including “This Vessel Does Not Exist” in the news section of issue 299! “This Vessel Does Not Exist” uses Machine Learning (ML) to generate images of “fake” ceramics. This Vessel Does Not Exist. Ceramic Review Magazine
Generating new hybrid vessels using machine learning (StyleGAN) trained on a large collection of photos of “vases” spanning different eras, cultures, and materials. https://thisvesseldoesnotexist.com
New temporary studio. Very fancy, but rent is super expensive 🙁 After a few million attempts some interesting results..