Today, the Princeton Vision & Learning Lab has made accessible, promising “Infinite photorealistic generative worlds“, all based on our favorite tools for deep2: Blender and geo-nodes. Plus an interesting fact, they are proud to be free from AI! This boosts our hopes for our project sky-high, pleases us very much and has stolen our Sunday walk, as we have been trying to install this on our equipment. Will report soon! ΓΌ
Just for fun I used our toolkit today for doing some small exercises designing nostalgic science fiction gear, constructions kits and wrist watches.
At the moment we learn a lot about the use of so called geo-nodes from the CAD-package Blender. They are very useful if you want to create automatically parallel solutions around variations of complex geometry. Generative design approaches permit to change a design concept through tuning variables. These may be diameters, with, length, positions, materials, etc. This quickly opens up an overwhelming space of potentialities, and the resulting problem is human choice. For this process, the user should define a maximal and a minimal bracket value for his variables. The user decides how many proposals he/she wants, and a program...
Created some hypernetwork-library for image creation today, mixing the work of a famous Swiss jeweler with injection molded kits for airplane models. So these are some images from the assembly toy library, and below some of the resulting AI-based jewelry.
Welcome to Shutian Gong, our new webmaster. She is a student at Nagaoka Institute of Design and will be supporting us in our communication efforts. So far, she already has found historical calligraphic alternatives for the creation of our final logo.
This morning, I have been hunting the web for fifty pictures of plastic model assembly kits for cars and guns and planes, all that childish stuff. Now I will create a so called hypernetwork library for some AI-based picture generation. It will take several hours for the PC to transform the pictures towards an intelligent library. I plan to tell the program to generate random images of alternative kits according to my own verbal descriptions, so that this output may surprise me. I will show these here tomorrow, I hope.
For a first demonstration of the deep2 concept during my upcoming design-talk, I am trying now to assemble my dream team; So far I have been able to convince a superb jeweler friend from Switzerland to help me as a critical partner and test user, and I have found on the fiverr-site an expert to create a first add-on in exchange with us. I still am searching for a suitable 3D metal printing service ready for an open-minded research collaboration with deep2. My job definition for the programmer:
