PLAY PODCASTS
Client-side AI in JavaScript using Tensorflow.js (gpn19)

Client-side AI in JavaScript using Tensorflow.js (gpn19)

Chaos Computer Club - archive feed · Mathias Burger

May 31, 201937m 0s

Audio is streamed directly from the publisher (cdn.media.ccc.de) as published in their RSS feed. Play Podcasts does not host this file. Rights-holders can request removal through the copyright & takedown page.

Show Notes

We will learn how to build a gesture classifier for a browser game using Tensorflow.js. Furthermore, we will go into the details of why MobileNets are so efficient and how to use transfer learning to reduce the amount of necessary training data. In a practical example I will show you how to build a gesture classifier for a vertical scrolling browser game. Along the way, we will discuss the Tensorflow.js API and a specific model architecture called MobileNet. In practice, edge computing is becoming evermore important and workloads are moved away from the backend. Tensorflow.js enables you to run AI models completely in the browser, which is also great for data privacy because often there is no more need to transfer sensitive data to backend servers. about this event: https://pretalx.entropia.de/gpn19/talk/ZHNT9Q/

Topics

gpn19108GPNgpn19GulaschProgrammierNachtKarlsruheErrorEntropia