TensorFlow.js - Bringing ML and Linear Algebra to Node.js
summary
TensorFlow.js brings training and inference to Node and the browser by running models with WebGL in the client and a TensorFlow C binding on the server, plus an alpha WASM backend for CPU. Pretrained packs like COCO-SSD object detector load in a few lines; MobileNet runs at 20 ms MobileNet inference (~50 FPS) in examples. The converter and bindings let about 200 core ops run natively and expose up to ~1000 TensorFlow ops, and browser Teachable Machine retraining actually retrains models live and saves usable TFJS artifacts.
AI-written summary · about
chapters (30)
- 0:00Introduction
- 1:04ML Terminology
- 5:10Why ML is taking off
- 6:42What is TensorFlow
- 7:18TensorFlowjs
- 11:29Why is this a good idea
- 15:22Platform support
- 15:58ML models
- 17:03Demo
- 19:33API
- 21:44Models
- 22:44Python Model
- 23:51Mobile Performance
- 24:59Code
- 25:19Highlevel API
- 27:18Train the model
- 28:46Make predictions
- 29:15Estimate function parameters
- 29:41Estimate function variables
- 30:35Chain function
- 31:26Tenderflow
- 32:12Graph
- 33:10Near Forum
- 33:48NodeRED
- 34:16LoScient
- 34:50Closing
- 35:17Resources
- 35:46Google Code Labs
- 36:10TensorFlow Textbook
- 36:40Thank you
related talks
20mTensorFlow.js Bringing Machine Learning to the Web and Beyond
24mTensorFlow.js + Node
41mTensorFlow.js 101: ML in the Browser and Beyond
31mIntroduction to Machine Learning in Tensorflow.js
48mTensorFlow.js
31mDeep Learning at the Edge with TensorFlow.js
28mMachine Learning on the web with TensorFlow.js
21mA Journey into the World of Machine Learning with TensorFlow.js