TensorFlow.js: Make a smart webcam in JS with a pre-trained ML model
summary
Step‑by‑step live coding builds a working smart webcam using TensorFlow.js and the pre‑trained COCO‑SSD model that detects about 80 object classes in real time, calls model.detect on the video stream, draws bounding boxes and labels, and counts items per frame while using requestAnimationFrame for smooth updates. The workshop remixes a Glitch codelab, shows a pet‑cam demo that triggers alerts, and discusses WebGL/WebAssembly acceleration and on‑device privacy so the demo actually runs live in the browser.
AI-written summary · about
chapters (10)
- 0:00Start
- 5:41What is TensorFlow.js?
- 11:31Pre-trained models
- 18:13Populate HTML Skeleton
- 21:48Adding Style
- 22:00Copy code to style.css
- 24:09Creating a JavaScript skeleton
- 33:11Using a TensorFlow.js machine learning model
- 33:25Copy code in to script.js
- 47:51Congratulations! What's next?
related talks
41mTensorFlow.js 101: ML in the Browser and Beyond
34mA Powerful Introduction to Machine Learning with Js
22mVue + TensorFlow.js: Building Your First ML Powered App
34mA Powerful Introduction to Machine Learning with Js
49mReal Time Face Detection through a Webcam Using AI
26mMachine Learning and AI for Angular Developers
34mMachine learning for front-end developers
32mWe Also Can Do It! Machine Learning in Javascript!