Bug 431831

Summary:	How much data gets sent to digikam.org for face recognition
Product:	[Applications] digikam	Reporter:	Samuel <spamarooo>
Component:	Faces-Recognition	Assignee:	Digikam Developers <digikam-bugs-null>
Status:	RESOLVED WORKSFORME
Severity:	normal	CC:	caulier.gilles, metzpinguin
Priority:	NOR
Version:	7.2.0
Target Milestone:	---
Platform:	Other
OS:	All
Latest Commit:		Version Fixed In:	7.2.0
Sentry Crash Report:

Description Samuel 2021-01-19 23:20:53 UTC

I use Digikam because I got tired of providing my data to the big privacy thieves of the internet (Google & Facebook) and I'm wondering just how much of my data is being sent to Digikam?

I've been using the face recognition in Digikam and it's been working well in the last update, but how much data is being sent to improve the engine, or collect any other data?

I have tried googling for this, searching all the "Faces-*" components of the forum, and no luck! Would be nice if you could update the website with the information, too.. I am sure many of us use this software as an alternative to the privacy thieves of the internet! I think we all would want to know just how much data gets sent back to Digikam servers, etc.

Comment 1 caulier.gilles 2021-01-20 03:35:12 UTC

Hi,

Q : How much data are send to a remote server as digikam.org or something like that to run faces management ?

Nothing...

All is done locally. this is the goal of this project. We don't use any external resources to complete face detection or recognition using IA.

You want to be sure. It's simple... before to scan your collections for faces, disconnect the computer from the network. You will see that all works as well...

We are open source, not Google.

Best 

Gilles Caulier

Comment 2 Samuel 2021-01-20 21:06:38 UTC

(In reply to caulier.gilles from comment #1)
> Hi,
> 
> Q : How much data are send to a remote server as digikam.org or something
> like that to run faces management ?
> 
> Nothing...
> 
> All is done locally. this is the goal of this project. We don't use any
> external resources to complete face detection or recognition using IA.
> 
> You want to be sure. It's simple... before to scan your collections for
> faces, disconnect the computer from the network. You will see that all works
> as well...
> 
> We are open source, not Google.
> 
> Best 
> 
> Gilles Caulier

Thank you for that! I have tried, and you're right - it works even without internet. But then I have to ask, how does the facial recognition algorithm improve over time if none of the faces in our catalogs are sent to Digikam servers? Do you just rely on 3rd party datasets, etc?

Comment 3 Maik Qualmann 2021-01-21 06:48:34 UTC

The face recognition is based on Deep Learning with Deep Neural Networks modules from OpenCV (a few terms to google). Of course, this algorithm must know what faces are. To do this, we download publicly available ready-made model data from a server of the KDE community. This happens once if the model data was not found on your computer, as I said download, no upload. The algorithm works completely locally on your computer. Now you give the faces names, so that their face vectors are saved in a database on your computer (recognition.db). In this way the algorithm learns and can recognize your faces in the future. Further improvements will result if we correct and optimize our code in new digiKam versions. Of course, we only test digiKam on our own image collections. Our hobby is photography too, and we have tens of thousands or more images on our hard drive.

Maik

Comment 4 Maik Qualmann 2021-01-23 22:08:37 UTC

I think it's all said. We do not send any data to our "digiKam server". I close the bug.

Maik