HI and AI
summary
Built an opt-in My Profile Tour that photographs visitors, runs face detection and landmark nets (initially OpenCV then Dlib), encodes faces and objects into vectors with VGG16 and matches them via PCA and graph routing, then prints individualized tours and problematic attribute labels from labeled faces in the wild. Uses Python with browser demos in ml5.js/p5.js. It sparks conversation and exposes bias (only four ethnicity labels), privacy friction (wall-of-terms unread, 30-day deletion promise), and parity with Google Arts & Culture.
AI-written summary · about
chapters (6)
- 0:00Start
- 3:43How is this relevant?
- 7:41My Profile Tour
- 16:22Lessons Learned
- 18:53Taking a wider view
- 20:35opportunity
related talks
32mSelf-Portraits in JavaScript
21mBring AI-Based Search to Your Web App
57mFoundations of Facial Recognition
49mReal Time Face Detection through a Webcam Using AI
23mWebAI: Building the next user experience in the browser
24mImage recognition and camera positioning with OpenCV. A tourist guide application.
52mAI-Powered Gamification for the Web
51mRealTime Exoskeleton-Mapping and Body Segmentation on the Browser with Tensorflow.js