Explore the potential of machine learning in JavaScript using TensorFlow.js in this 26-minute conference talk. Gain inspiration from creative prototypes and learn how to implement computer vision in the browser. Discover how to use ML anywhere JavaScript runs, including retraining and writing models. Delve into object recognition, face mesh, and body segmentation techniques. See practical applications like invisibility cloaking and teleportation effects. Learn about combining ML with web technologies for clothing size estimation. Explore transfer learning with Teachable Machine and Cloud AutoML. Understand TensorFlow.js APIs and model inference performance improvements. Get insights on client-side and server-side benefits of using TensorFlow.js. By the end, acquire the knowledge to start your own ML projects using JavaScript.
Overview
Syllabus
Intro
Use ML anywhere JavaScript can run
Run, Retrain, Write Reuse existing models, or create your own
For anything you may dream up
Object Recognition
Face Mesh
Body Segmentation Distinguish 24 body areas, across multiple bodies, real time.
With a little bit of creativity...
Invisibility Made in just 1 day of coding
Teleportation
Combine with other web tech Such as clothing size estimation
Teachable Machine Transfer Learning
Cloud Auto ML
Code
TensorFlow.js APIs
Model Inference Performance Only Server
Hugging Face DistilBERT 2x Perf boost using Node.js
3 client side super powers
Server side / Node.js benefits That make TensorFlow.js easy to use
What will you make?
Taught by
TensorFlow
