Machine Learning Magic for Your JavaScript Application
summary
TensorFlow.js brings training and inference to browser and Node.js with concrete claims: pre-trained models like BodyPix and a speech-commands model run via WebGL, and the Universal Sentence Encoder produces 512-dimensional embeddings. An intent classifier trains from ~6,000 Snips sentences in ~20 seconds on CPU; a bi-directional LSTM tagger takes 10–20 minutes and exports JSON+binary model files. Demos include a webcam-driven Pac‑Man in a WeChat mini-app, AirBnB client-side sensitive-photo checks, Uber Manifold numeric acceleration ~100x, and Magenta Studio music plugins.
AI-written summary · about
chapters (11)
- 0:00Introduction to TensorFlow.js
- 3:24Using pre-trained models
- 6:16Speech commands model demo
- 8:40Introduction to training
- 10:46Training a text classifier
- 16:28Deploying the custom model
- 21:31Advanced model training
- 25:58Tagging model demonstration
- 30:41Production best practices
- 32:01TensorFlow.js applications
- 37:29Getting started resources
related talks
41mTensorFlow.js 101: ML in the Browser and Beyond
15mMachine learning for next gen web apps with TensorFlow.js
31mIntroduction to Machine Learning in Tensorflow.js
20mTensorFlow.js Bringing Machine Learning to the Web and Beyond
37mMachine Learning Principles, with JavaScript?
57mMachine Learning with JavaScript
54mAdventures in Javascript Machine Learning
29mSuperpowers for next gen web apps: Machine learning