2019
Bachelor Thesis
Art(ificial)
Developing an Pixel-to-CNC algorithm & implementation for painting AI generated art with an industrial robot.
Summary
Artificial
An autonomous installation consisting of an industrial robot painting AI generated art on canvas.
Abstract
This bachelor thesis consists of a mostly autonomous industrial robot, painting AI-generated art. Various pre-trained GAN models were used to generate the ground truth artworks which were then converted into cartesian coordinates with a novel algorithm. The 1000 lb system is capable of painting with acrylic color on surfaces such as canvas, paper, or stone. It attempts to emulate a defining aspect of human intelligence: artistic creativity. The machine painted various artworks for ten hours during a two-day exhibition without human intervention. This project was kindly supported by KUKA robotics.
Gallery
Close Summary
Abstract
SUPERvisor was a team effort of UC Berkeley students, made during the „Jacobs Institute COVID-19 Design Challenge“. During this three week hackathon, six members across three continents worked remotely on a two phase product for medical workers. The phase 1 model consists of an 3D-printable AR face-shield, enabled by a smartphone and $10 of material. Our team conducted user research with medical workers and military personell in the US, UK, and Germany. The project has been awarded the 2020 Linda Tischler Award by Fast Company.
Gallery
Close Summary
MicroMix
Making Data Science Intuitive for Genomic Research.
Abstract
A visual programming application for filtering, merging, transforming, and visualizing large genomic datasets. Programmable forms and logical and/or gates can be added, stacked, and customized. By combining filters and data transformations, researchers can easily perform complex data science without any coding experience. An additional novel WebGL 3D heat map, offers high-performance visualizations of large datasets in 2D and 3D, seamlessly integrated in MicroMix. The system is build on a Vue.js frontend, Python backend, and a MongoDB database. Python's powerful Pandas and NumPy libraries enable efficient computations even with large datasets. Sessions are stored in the highly efficient Apache Parquet format. This project has been developed at the Helmholtz Research Society by me and supervised by Jun. Prof. Lars Barquist and Prof. Jörg Vogel.
Gallery
Close Summary
2019
Machine Learning
Machines describing man
Generating novel art with Generative Adversarial Networks.
Summary
Machines describing man
Generating novel art with Generative Adversarial Networks.
Abstract
Generating art with Machine Learning. This project revolves around the ability of computers to dream up portraits and abstract art. It documents the process of finding datasets, preparing them with high precision, benchmarking hardware and finally training multiple Generative Adversarial Networks (Karras et al.) on different genres of 2D art. It acts as a first attempt of proving that machines gain skills that have been previously reserved to truly intelligent species.
Gallery
Close Summary
2018
Tangible Interfaces
This is (not) a drill
A tangible interface for displaying the change of the polar ice cap over the past 40 years.
Summary
This is (not) a drill
A tangible interface for displaying the change of the polar ice cap over the past 40 years.
Abstract
Interactive installation showing the extent of the polar ice cap from 1979 to 2018 featuring a 1.8 meter tall ice core sculpture. The projects goal was to compress accurate satellite data from NASA into an approachable visualization and creating an input system, that is closely tied to real scientifc methods, used for analyzing ice cores.
Gallery
Close Summary
Abstract
Stud.io is a projector that transports the collaborative spirit of art studio into your own home. Working, learning, and existing remotely is hard. stud.io as a concept stems from the motivation to address this problem by introducing tangible interactions into a remote creative workflow. Bring stud.io into your home: get instructions from experts, work next to peers, and follow along in real-time.
Links
Gallery
