Thank you Andreas for this nice introduction and pointing out this error.

Yeah, before we talk about our online annotation tool for collaborative image analyzers like

Google Docs, thank you Jenny for the hint.

The mandatory PIS slide.

I started back in April last year, that's my thought, PIS, two first author contributions

for conferences, one journal is still under review since August last year and we have

two papers under preparation, one for the MICHI and one about EXACT and now I want to

talk a little bit more about that one.

So let's see first a small introduction.

In this lab we have a slide runner, a very good annotation tool for whole slide images

and now we wanted to make the next step, make it wider accessible and make it easy to collaborate

on image annotation.

So we take a look in the literature and said okay, we have two main features we need to

implement to do that.

First, it has to be online so you can work on it with your mobile phone, tablet or some

big screen like Sven has it and we have two key features.

First, data privacy, meaning that the data set is staying in your house or in your institute

or in your hospital but you can share it without copying all the meta information, we will

look about that later and we have deep learning support, that's our two key contributions.

Also we needed a REST API to synchronize with existing applications, real time cooperation,

you can in real time see what other people are annotating and you can say okay, I want

a second opinion for that area, that's in the medical field quite important.

We support crowd sourcing and crowd algorithm collaboration, so if you have already a trained

model you can upload your annotations and users can correct and enhance them.

Also we have plugin support with just a few lines of HTML code and Python, you can write

your own plugin scripts and say okay, we want to extend the server to this new feature.

Also we have some deep learning support but I want to talk about now into more detail.

So let's start, we are an informatics lab so let's talk a little bit about software

architecture.

First, we have a user who wants to annotate data and normally our users are not informaticians

so we make it as easy as possible to annotate data for them, first by using a browser and

if you want to use it on your local computer, you dock out everything.

So basically first we have engine X as a reverse proxy, then we have Django, Django is our

Python server with GuniCorn for load balancing and handling Django instances, then we have

Postgres database to store the annotations and well you can save images on your local

hardware.

So if you want to use Xact on your own computer you have to do two things, first check out

the GitHub repository, it's the last slide, the FNQR code to make it easier for you to

follow that link and just if you have installed Docker, Docker compose, we have for production

or for development, build files, you just execute that build files and you can use the

Xact server on your system independent if you have a Linux or a Windows system, maybe

Mac but I don't have it.

So and now you have not five to ten users, so you have a few hundred users, with our

Dockerization system we made it scalable so you just launch more instances, you go to

a centralized database and hard drive and you can say okay we scale up the system with

our demands, with the user demands.

So now let's come to the features, the idea was to streamline the annotation process and

to standardize the annotation process.

We had a lot of problems with annotation by multiple experts because one says okay it's

a mitotic figure, he added the class mitotic figure, the next one says it's mitosis, that's