Bug 435367

Summary:	"deep face" a free source code for better face detection
Product:	[Applications] digikam	Reporter:	aviv <avivkuh8>
Component:	Faces-Detection	Assignee:	Digikam Developers <digikam-bugs-null>
Status:	RESOLVED FIXED
Severity:	wishlist	CC:	caulier.gilles, michael_miller
Priority:	NOR
Version First Reported In:	7.2.0
Target Milestone:	---
Platform:	Microsoft Windows
OS:	Microsoft Windows
Latest Commit:		Version Fixed/Implemented In:	8.5.0
Sentry Crash Report:

Description aviv 2021-04-04 22:39:20 UTC

I thought of a great idea to make face recognition in software significantly more realistic

In GitHub there is an open source project called "deep face" that a lot of companies use

And is at an almost human level of accuracy (95 percent success and maybe more) it can be a great idea to incorporate it within this amazing software and thus make face recognition work well with AI capabilities

Project link:
https://github.com/serengil/deepface

Comment 1 caulier.gilles 2021-04-05 03:36:40 UTC

Deep Face is Python library. We use C++, not Python.

We take choice of OpenCV library for that...

Mixing C++ and Python which make a mess in the project, especially for the core implementation.

Gilles Caulier

Comment 2 caulier.gilles 2024-10-06 16:25:34 UTC

The new YuNet DNN model is now the default one for the face detection in 8.5.0 which give excelent and fast results.

Comment 3 Michael Miller 2024-10-15 22:25:55 UTC

Hello Aviv,
I'm familiar with DeepFace.  DeepFace is a collection of different face detection and face recognition models.  If you look into the source code of DeepFace, you'll see YuNet (detection) and SFace (recognition) in the list of models DeepFace supports.  digiKam 8.5.0 includes both YuNet and SFace.  Many of the other models in DeepFace are older and slower. 

In different words, digiKam is now using 2 of the newest and most accurate models used by DeepFace.  We think this will be a big improvement for users.

Cheers,
Mike