1

Currently using Adobe Lightroom CC 2019 Classic to organize (tag) 40.000 jpg family photos.

However, the built in face-recognition is terrible as it only recognises ~40% of the photos with faces and only ~30% of the people correctly. So I'm searching an alternative for the face-recognition.

Can someone recommend software that does face recognition with high accuracy? To tag people in jpg photos.

Google's cloud API service Vision is very good with objects and I'm curious about user experience with face recognition. Or machine learning based models with tensor flow look promising too. Unfortunately I have not much insight into current face tagging software what's available today, so maybe someone can recommend or point in the right direction?

6
  • 3
    "The best" always depends on the criteria that are applied. Every software somehow thinks it is the best. See also this meta discussion Apr 17, 2020 at 15:52
  • I am searching a drastically more accurate face recognition software than Lightroom. Are the cloud APIs (e.g. Google's Vision) performing better and are the way to go? I'm not in the field of photo editing / face recognition, so I'm asking what the state of the art, currently is. Im not interested in philosophical discussions on phrase, but in solutions and thoughts for a concrete problem. Apr 17, 2020 at 17:00
  • 1
    When would you be satisfied? 50%? 80%? 90%? 99%? The latter is still off by 400 of 40.000 pictures. Apr 17, 2020 at 20:32
  • "I'm asking what the state of the art, currently is." - it all depends on money. If you spend 1B US$, you can probably buy the best face recognition software the world has to offer. Maybe you need to buy a new PC too in order to run it. So yeah, you may think it's just about phrases, but it's really not. It's more about knowing what you really want and what you can afford. Apr 17, 2020 at 20:37
  • 1
    Nobody knows how much money you're willing to pay. We can imagine that you're willing to pay 12 bucks per month, which is the same price as LR CC 2019. But maybe you would even spend 2000 bucks instead of manually tagging 20.000 pictures, which might take you days to complete. We don't know. Apr 17, 2020 at 21:22

1 Answer 1

1

Here's two I've used with reasonable success. You can try them and see if they do any better for you than Lightroom does:

  1. TagThatPhoto: tagthatphoto.com - subscription $19 to $59 USD per year, but you can try it for free. They use their own "patented face recognition technology to automate the task of organizing your photos."

  2. digiKam: www.digikam.org - free and open source. Uses a "Faces database to store recognition vectors from neural network processing based on OpenCV".

DigiKam is a full-featured photo management system, but is more complex and harder to use than TagThatPhoto. I started off using TagThatPhoto, but then switched to DigiKam especially for its geolocation features which TagThatPhoto doesn't have.

4
  • while digikam can detect faces, recognizing them was a huge fail for me. People faces that it thought matched were totally different. 90% false positives. Increase the accuracy requirement only results in missed matched.
    – cybernard
    Jan 25 at 18:37
  • @cybernard - Was that before or after the face recognition improvements they added in Version 7.2, March 2021?
    – lkessler
    Jan 26 at 5:53
  • This is for the newest version download December 2021.
    – cybernard
    Jan 26 at 16:29
  • Tagthatphoto has GUI issues where the unknown group has no progress bar, and therefore no way to go back to where you left off. A major pain when you have 20,000+ faces.
    – cybernard
    Feb 28 at 21:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.