Meetup 2019.1: Tensorflow.js and Stereoscopic Real-Time Style Transfer

Dear Meetup members, we're glad to announce the next meetup on Wednesday, February 20th, beginning at 6:30pm at TNG Technology Consulting GmbH, Arabellastr. 4a. See you there!

Talk 1

Title
Client-side Artificial Intelligence in JavaScript using Tensorflow.js

Abstract
Traditionally, AI models are trained and evaluated in specific
high-performance or cloud environments. Currently, there is a trend
towards Edge Computing which is primarily caused by better processors
and hardware acceleration, especially on mobile devices. Furthermore,
data protection is becoming ever more important and running AI on the
client-side, without data leaving the device, becomes increasingly relevant.

With Tensorflow.js models can be trained and evaluated in the browser or
in the backend. This enables web applications to use AI online as well
as offline. Moreover there is vendor independent hardware acceleration
with WebGL when using the browser. That means no lock-in on CUDA and
better performance than running solely on the CPU.

In a practical example the complete workflow for developing a gesture
classifier will be presented. Participants will gain insights into the
Tensorflow.js API and learn how to apply Transfer Learning. For
demonstrative purposes the classifier will be integrated into a vertical
scrolling airplane game. Training and evaluation will be done completely
in the browser so that no data leaves the device.

Example
Vertical scrolling plane shooter controlled by a gesture classifier.
Animated gif of the practical example:
https://home.tngtech.com/~burgerm/skyfire_using_gesture_classifier_small.gif

Short Bio
Mathias Burger is enthusiastic about open source software and always
interested in the latest trends in IT, especially in computer vision. He
is a Senior Consultant at TNG Technology GmbH and is currently working
on a defect detection system in the aerospace industry.

Talk 2

Title
Stereoscopic Real-time Style Transfer AI - Art is not what you see?

What if Virtual Reality glasses could transform your environment into a three-dimensional work of art in realtime? What if every detail of the world would be shown in the style of a painting from Van Gogh? One of the many interesting developments in the field of Deep Learning is the so called "Style Transfer". It describes a possibility to create a patchwork (or pastiche) from two images. While one of these images defines the the artistic style of the result picture, the other one is used for extracting the image content. The Hardware Hacking Team from TNG Technology Consulting managed to build an application using OpenCV and Tensorflow to realize such goggles. When you see the world through these glasses, your environment will be displayed in the styles of famous painters like Claude Monet or Pablo Picasso. Within this talk you will be introduced into the scientific field of Realtime Style Transfer. It will also cover and explain in detail the Deep Learning techniques used for this application.

Short Bio
Speaker 1: In his role as an Associate Partner for TNG Technology Consulting in Munich, Thomas Endres works as an IT consultant. Besides his normal work for the company and the customers he is creating various prototypes - like a telepresence robotics system with which you can see reality through the eyes of a robot, or an Augmented Reality AI that shows the world from the perspective of an artist. He is working on various applications in the fields of AR/VR, AI and gesture control, putting them to use e.g. in autonomous or gesture controlled drones.
Speaker 2: Martin Förtsch is an IT-consultant of TNG Technology Consulting GmbH based in Unterföhring near Munich who studied computer sciences. Workwise his focus areas are Agile Development (mainly) in Java, Search Engine Technologies, Information Retrieval and Databases.