UVM Art + AI Research Group Phase 2 (2021)

Submission Number: 99
Submission ID: 145
Submission UUID: 582e0882-fb21-435c-b76c-2e8eee6f0387
Submission URI: /form/project

Created: Wed, 05/05/2021 - 11:00
Completed: Wed, 05/05/2021 - 11:18
Changed: Thu, 05/05/2022 - 03:37

Remote IP address: 24.2.153.15
Submitted by: Jennifer Karson
Language: English

Is draft: No
Webform: Project
UVM Art + AI Research Group Phase 2 (2021)
Northeast
Sample Step 3280.jpg
ai (271), cuda (222), python (69)
Complete

Project Leader

Jennifer Karson
802-238-7768
{Empty}

Project Personnel

Keri Toksu
Clement Fisher, Halina Vercessi-Clarke, Cameron Bremner, Sydney Culbert
{Empty}

Project Information

The UVM Art + AI Research Group is seeking artist-programmers who have a solid understanding of geometry, coding skills in Python and Processing, and who have the interest and ability to work with digital visual tools such as Adobe Creative Suite (education will be provided). These team members will: help to optimize our image generation on the VACC; design methods of displaying the images for exhibition; digitize artworks to expand our dataset; and assist in defining genetic traits within our existing genetic algorithm. Image generation will be optimized by exploring how latent vectors are used in StyleGAN2-ADA. The genetic traits we wish to develop further include grid variables such as point, line and plane; micro and macro spatial relationships; and color palette. We are excited to welcome these new artist-researchers into our research group and they will participate in our team bi-weekly meetings where they can expect mentoring, education and hands-on learning in the realm of machine learning.

Project Information Subsection

1. Image production on the VACC using unique artist-made dataset, to generate 2D and 3D artifacts toward a goal of fabricating 3d artifacts in addition to 2d artifacts.
2. Document experimentation with StyleGAN2 and StyleGAN2 ADA on Model Cards, tracking GPU usage, image generation, etc.
3. Contribute to the publishing of images through art exhibitions.
4. Support programming our research group's existing genetic algorithm with new genetic traits.
5. Contribute to the group's understanding of latent space and latent space walk videos by identifying analogies (beyond mathematics) that allow the group, (as well as an audience of artists and art historians) to understand latent space as it serves as a canvas for image making.
{Empty}
Undergraduate or graduate students.
{Empty}
Interest and experience in programming and the arts.
{Empty}
University of Vermont
{Empty}
NE-University of Vermont
06/14/2021
No
Already behind3Start date is flexible
June - August 2021
{Empty}
{Empty}
{Empty}
  • Milestone Title: Understanding of Latent Space / Exploration of Genetic Algorithm / Digitization of Artwork
    Milestone Description: • Understanding of how StyleGAN2-ADA uses latent space and how latent vectors are used to generate images.
    • Exploration of possibilities to control genetic traits based on color palette.
    • Digitization of artworks for two new training sets.
    Completion Date Goal: 2021-11-15
  • Milestone Title: Experimentation with Latent Space / Programming of Genetic Traits / Digitization of Artwork
    Milestone Description: • Observation of the effects of modifying latent vectors and different StyleGAN2-ADA parameters on image creation.
    • Programming of new genetic traits, based on color palette, for the genetic algorithm.
    • Digitization of artworks for two new training sets.
    Completion Date Goal: 2021-11-15
  • Milestone Title: Intentionality in Image Production / Dataset Production / Exhibition Model
    Milestone Description: • Pinpoint the regions of latent space that result in the most visually appealing images and use the corresponding latent vectors to generate the best images.
    • Provide sample images for new datasets.
    • Provide model for exhibition.
    Completion Date Goal: 2021-11-15
{Empty}
We will share image artifacts and we are developing educational tools for artists.
Exhibitions and Educational programs in under development.
Students will learn how to work with and program machine learning platforms from direct hands-on experience. Through working and sharing information with our existing team, they will be immersed in StyleGAN research on the VACC. Students will also learn of methods employed in a contemporary art practice, as well as methods of visual research and ethnographic inquiry.
{Empty}
We are excited to share our aesthetic and ethnographic research with the Cyberteam. We are also developing an educational program that may be of interest and perhaps might contribute to broader educational efforts by the Cyberteam community.
Use of the VACC's DeepGreen cluster.
{Empty}

Final Report

{Empty}
{Empty}
{Empty}
{Empty}
{Empty}
{Empty}
{Empty}
{Empty}
{Empty}
{Empty}