Abstract 3D Model Generation Using Generative Adversarial Networks and Mandlebulbs
Mentor 1
Christopher Willey
Start Date
16-4-2021 12:00 AM
Description
This research project is based on generated images and how they can relate to each other in 3D space. The goal of this project is to take 2D layer data from 3D Mandelbulbs, have a machine learning program generate new images. Then take those generated images to generate 3D models that can be rendered, 3D printed, or manufactured via digital fabrication methods. The method for this research is the generation of images via a program called Mandelbulb3D. Training the newly created dataset in a python program called HyperGAN. Then taking the resulting images and convert the images into image stacks and export the stack as a voxel .obj files with a program called Fiji/ImageJ. The desired outcomes of this research are to determine whether using a machine learning algorithm to generate individual layers, then join them into a voxel model is an effective means for semi-procedural 3D model generation.
Abstract 3D Model Generation Using Generative Adversarial Networks and Mandlebulbs
This research project is based on generated images and how they can relate to each other in 3D space. The goal of this project is to take 2D layer data from 3D Mandelbulbs, have a machine learning program generate new images. Then take those generated images to generate 3D models that can be rendered, 3D printed, or manufactured via digital fabrication methods. The method for this research is the generation of images via a program called Mandelbulb3D. Training the newly created dataset in a python program called HyperGAN. Then taking the resulting images and convert the images into image stacks and export the stack as a voxel .obj files with a program called Fiji/ImageJ. The desired outcomes of this research are to determine whether using a machine learning algorithm to generate individual layers, then join them into a voxel model is an effective means for semi-procedural 3D model generation.
