Unlocking the Power of ML for Your JavaScript Applications with TensorFlow.js

Explore the TensorFlow.js ecosystem in this 44-minute conference talk from TF World '19. Learn how to integrate existing machine learning models into JavaScript applications, retrain models using custom data, and extend beyond browser environments to other JavaScript platforms. Witness live demonstrations of innovative models and featured applications, including L'Oreal's WeChat virtual try-on plugin and Mila's radiology diagnostic tool. Discover multiple entry points for leveraging TensorFlow.js, understand its advantages, and delve into practical applications such as pose detection, object detection, lip sync, and virtual makeup try-on. Gain insights into future directions, including hair color prediction, and learn about technical challenges and React Native support. Presented by Kangyi Zhang, Brijesh Krishnaswami, Joseph Paul Cohen, and Brendan Duke, this talk offers valuable knowledge for JavaScript developers looking to harness the power of machine learning in their applications.

Intro
Outline
TensorFlow.js Multiple starting points for your needs
Why TensorFlow.js?
Pose Detection
Object Detection
Application - Lip Sync
Application - Virtual Makeup Try-on
How TensorFlow.js fit the bill
Future directions Hair colour
Technical Challenges
React Native support
Joseph Paul Cohen, PhD