128 numbers
Radical Networks 2019 Proposal
Format: [Workshop]
Name: [Aarati Akkapeddi]
Pronouns: [She/Her/Hers]
Location: [NYC, NY]
Email: [aarati.akkapeddi@gmail.com]
Twitter: [@AaratiAkkapeddi]
Repo: [https://github.com/AaratiAkkapeddi/face2map]
Url(s): [http://aarati.me/project.html?project=project-encoding]
Consent to being photographed?: [Yes]
Consent to being on the livestream?: [Yes]
Speaker Bio and Profile Picture
[Aarati Akkapeddi (b. 1992, Edison, New Jersey) is an Indian-American cross-disciplinary artist interested in the poetics and politics of datasets. She works with both personal and institutional archives to explore the ways in which identities and histories are shaped by different ways of collecting, preserving and presenting data. Currently, she is investigating the relationship between archival ethics and the creation of image datasets used in machine learning. Aarati has exhibited at venues such as The Java Project in Brooklyn and The Irregulars Art Fair in New Delhi. She is also an adjunct faculty at The New School.]
[I identify as being a woman of color in tech. I identify as being desi. I identify with the mission of organizations such as AI Now in shedding light on algorithmic bias and algorithmic censorship. All of these communities are important in the work that I do because they allow me to consider the complexities of ethics and technology in different ways. For instance, I do not uphold open-source/transparency as a universal solution as I feel that issues of ownership, appropriation, and cultural heritage call for a more nuanced approach.]
Note: Format for bio picture should be PNG/JPG/GIF, 256x256px.
Description
[This workshop is about facial recognition. The title 128 numbers, comes from the 128 measurements of the face used in Google's FaceNet neural network to identify faces. Because it is a blackbox algorithm, although 128 numbers are assigned to a face, almost like a unique fingerprint, the correspondance between the facial features of a person and these numbers is unknown. This was the basis of my own work (Encoding)[http://aarati.me/project.html?project=project-encoding], where I took family photographs and visualized their corresponding 128-dimensional vectors in the form of topographic maps. In this workshop we will go over how to use python to first identify faces from photographs, and then we will explore different ways to visualize the 128-dimensional vectors. The goal of this workshop is not only to introduce facial recognition from a technical standpoint, but also from a conceptual one as well, discussing the ethical implications of faces as data.]
Length: [3 hours]
Workshop technical requirements and materials list (if applicable)
Equipment/technical requirements: [Projector, WiFi, Extension chords/outlets, table space]
Maximum number of attendees: [12]
Full list of materials needed for attendees: [laptop, Sublime Text (or preferred code editor), Python installed if not on a mac]
Additional Info / Links / References
[https://www.cv-foundation.org/openaccess/content_cvpr_2015/app/1A_089.pdf]