Mask || No Mask? Detect Masks with ml5.js
summary
Client-side image classifier built with ml5.js (MobileNet feature extractor) uses a webcam to collect labeled examples via two buttons, trains on roughly 20 images per class (author notes 20 worked for the demo but recommends 100+), displays training loss, then outputs mask / no mask labels with a confidence score. The model saves to the repo assets and plugs into a serverless Twilio Programmable Video room to run live; detection can fail for masks worn incorrectly and needs more data to improve.
AI-written summary · about
chapters (11)
- 0:00Introduction and setup
- 1:26What is ml5.js?
- 3:59Supervised vs unsupervised learning
- 6:29Classification vs regression
- 8:14Getting started with Twilio
- 10:00Training the mask model
- 12:32Adding training images
- 14:01Using the trained model
- 15:25Testing edge cases
- 15:47Summary and resources
- 16:52Q&A session
related talks
13mml5.js - Friendly machine learning for the web
32mWe Also Can Do It! Machine Learning in Javascript!
34mMachine learning for front-end developers
53mNodeJS, ML, K8s and Unethical Face Recognition
34mA Powerful Introduction to Machine Learning with Js
48mThe Future of Machine Learning & JavaScript
19mML at Client Side
49mReal Time Face Detection through a Webcam Using AI