Gender Tapestry - New ways of looking at gender classification
summary
Claim: gender maps to a 3D color space rather than a linear spectrum. Implementation: a Python Flask web app loads a TensorFlow multi‑label classifier trained on generated faces plus a public gender dataset, using six pronoun labels to output percentage scores. Numbers are converted from a numpy array into RGB via a quirky modulo/normalisation step, then applied to binary‑thresholded portraits to make light and dark versions. Results are consistent per image but shift with presentation; uploads feed the evolving mosaic and GAN and are uploaded images become training data.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:01Gender as 3D color space
- 3:53Gender and color in marketing
- 5:19Gender Tapestry project
- 7:05Binary gender in AI challenges
- 8:51Non-binary gender recognition
- 10:20AI training on artificial faces
- 11:45Bias in AI gender classification
- 13:29Gender Tapestry code
- 16:43Gender visualization mosaics
- 18:08Physical to web artwork
- 19:33AI gender perception bias
- 21:03Equitable AI and gender
- 22:28Future of AI and gender
- 23:31Q&A: Gender color
related talks
40mGender 101
22mHI and AI
56mGlobal Bias in AI: When Western Data Shapes the World
27mDesigning More Inclusive AI with More Inclusive Workflows
14mCSS VS. My Gender: A Study In Transitions
22mLet’s make the world a better place: Python for Social Good
39mBecoming Conscious of Our Unconscious Bias in Design
53mTranscendence: The Power of Representation