Deep Learning at the Edge with TensorFlow.js
summary
TensorFlow.js runs pre-trained models in browser or Node.js, using WebGL for GPU math and tfjs-node uses native TensorFlow C lib for faster server-side inferencing. Models must be TensorFlow or Keras formats via the tfjs converter; ONNX/PyTorch aren’t supported. Tensors are immutable, memory leaks appear if disposals aren’t applied and TF.tidy is necessary. Current op support is limited — ~200 ops supported versus TensorFlow’s ~800 — but practical builds exist like a PoseNet->MIDI video Theremin and Drone Aid SOS detector.
AI-written summary · about
chapters (11)
- 0:00Machine Learning with TensorFlow.js
- 0:53Machine Learning & AI
- 1:14Programming VS Machine Learning
- 1:54Artificial Neural
- 4:57TensorFlow.js is an open source library to train and deploy machine learning models with JavaScript
- 5:40Why JavaScript?
- 10:11Tensors
- 19:30instead of training how about
- 21:09Converter supports only TensorFlow and Keras models
- 22:00Model requirements
- 23:03Limited number of supported operations
related talks
31mIntroduction to Machine Learning in Tensorflow.js
20mTensorFlow.js Bringing Machine Learning to the Web and Beyond
24mTensorFlow.js + Node
37mTensorFlow.js - Bringing ML and Linear Algebra to Node.js
54mDeep Learning in JS
48mTensorFlow.js
32mDeep Learning in JS
28mMachine Learning on the web with TensorFlow.js