Data Visualization x Machine Learning
Workshop Supported by the SDN grant 2021
Machine Learning (ML) and Data Visualization (DV) are shaping multiple domains of the human experience, driven by the increasing offerings of different tools and applications that promote accessibility to these fields even for designers who do not have a background in computer science, making them able to consider a broader spectrum of methods and solutions during their work.
This workshop proposed a learning journey in which students, designers-artist and workers were enabled to create digital artifacts based on ML technologies and Data visualization techniques, through simplified workflows, a learning by doing approach, tools guided use and open resources.
It was open for anyone interested and curious to explore the application of Machine Learning and Data Visualization techniques in creative practices.
The workshop was held was held online for a total duration of 4 days, divided over two consecutive weekends, Oct. 22-23 and Oct. 29-30, so that students and workers could easily participate.
- The first part of the workshop experimented with the application of existing machine learning models and tools to analyses images, video and sound to produce annotated media. Here annotated media can be seen as media with a high-level description of its contents. These descriptions enabled complex media to be organized, mapped out and visualized in a more-or-less human vocabulary.
- The second topic of the workshop was working with tools to analyses and further simplify descriptions such that they can be visualized. The two subjects were addressed in a hands-on manner through instructions, exercises and examples. The workshop required a bit of programming, which was be done using the OPENRNDR framework.
Edwin Jakobs main developer of the OPENRNDR open source framework for creative coding, whose work revolves around software and data, ranging from visual art, information design to software engineering.
The participants, divided into four heterogeneous groups, worked with a dataset of images scraped from Know Your Meme, an Internet Meme Database. To analyze and visualize this data, they designed and developed four meme-explorer interfaces using the Openrndr framework and other ML tools.
Finding the different cultural stereotypes that are propagated through meme culture, using the inherent biases that are found in the models classification of image data.
Determining the potential of machine learning as a tool to identify the age of images, within large datasets, based solely on the visual content of the pictures themselves.
Visualizing the most present emotions and moods of society over time.
Analaziying categories of meme in terms of statistics such as views, favorites, classification of memes in terms of Gen XYZ, Logits.