Use Ai in your Website using Tensorflow.js
summary
Builds a convolutional neural net with Keras/TensorFlow, trains on the MNIST 70,000 images (28×28 grayscale) in Python, exports the model into a models folder and loads it with TensorFlow.js in the page. The demo runs entirely in the browser, processes canvas pixels converted to arrays, outputs 10-way digit probabilities (example shows ~99% confidence), and uses standard conv + dense layers; code and a live demo link (bit.ly/tfjsdemo) are provided.
AI-written summary · about
chapters (20)
- 0:00Introduction and speaker intro
- 1:21Why TensorFlow.js matters
- 3:25What is AI in practice
- 7:57Building a digit recognizer
- 11:46Code walkthrough and setup
- 17:05Live demo and results
- 18:51Further learning resources
- 20:03Q&A session begins
- 21:05How to train a dataset
- 21:49Hidden layer size decisions
- 22:54AI vs human brain comparison
- 24:20When to use backend processing
- 25:23Why AI feels intimidating
- 27:09Handling incorrect predictions
- 28:53AI JavaScript learning resources
- 31:18TensorFlow.js vs Python for AI
- 33:04Training models in JavaScript
- 33:49Other TensorFlow.js projects
- 35:15Current side projects
- 35:58Closing remarks and thanks
related talks
21mA Journey into the World of Machine Learning with TensorFlow.js
41mAI in Web: TensorFlow js
34mA Powerful Introduction to Machine Learning with Js
24mThe Future of Machine Learning & Javascript
54mDeep Learning in JS
35mHow to AI in JS?
28mMachine Learning on the web with TensorFlow.js
20mTensorFlow.js Bringing Machine Learning to the Web and Beyond