Facial Recognition / Additional Photos / Google Takeout

Hi - just starting to play around with PhotoStructure as this seems to be a great way to host my family’s pictures on a private web site. I’ve been using Google Photos to help build my initial collection. One of the key things I want to have is facial recognition, which I believe is coming with the 1.2 PS release.

I get that I can do a Takeout and use that to have PS read/use the people tags to build its initial database. Until 1.2 gets released, though, I’m guessing that I have to still upload the photos to Google to do the automatic person tagging (I will be uploading hundreds of pictures at a time, multiple times) and then somehow import those new photos into the PS library? If so, is the right way to do that to just export the new photos in a new Takeout file, add them to the directory I’m storing my photos in, and then rebuild the database? Or is there an easier way to do it? I’m currently at about 1500 photos, and I expect that number to increase to at least double that size.


I don’t know if you’ve thought about this, but there are other workflows (that don’t involve Google Photos) that work.

My process is this:

  1. Move photos from Phone to home network using Nextcloud (there are lots of other ways to do this as well)
  2. Once the photos are in the proper folder on my home network, they are picked up by Photostructure
  3. Simultaneously, TagThatPhoto scans that folder and finds faces. The TagThatPhoto UI suggests people to faces, which I confirm. Once that’s done, the tag is written to the photo file, and Photostructure picks up that tag on the next sync.

It will be nice when Photostructure does facial recognition so I can get rid of TagThatPhoto, but in the meantime the process is really pretty simple. And best of all, I don’t have to deal with Google Photos at all!

1 Like

Great suggestion, and $20/year isn’t a bad price! Unfortunately, I’m running on a Mac at home, and it looks like this is Windows only. I might be able to get around that running a VM, though.

Ah, yes, I should have specified Windows only.

Does anybody know if there are Mac-compatible desktop softwares that do facial recognition/tagging?

Make sure to vote for that feature request!

Meanwhile, if you’re looking for Mac software to manage tags (including face recognition), I recommend Digikam. It works on Mac, Windows and Linux. I use it to manage all tags on the photos that PS displays.

1 Like

I did not know about TagThatPhoto and just reviewed their web site. It looks like an interesting product but Windows only. My daily system is Mac and Linux but I do have a Windows system (and VM). I might explore TagThatPhoto just to see if it is worthwhile enough to use Windows for.

For now, I am using digiKam for the majority of my photo tagging and organization. digiKam is multi-platform and I am currently running it in Docker. digiKam also does facial recognition and its performance is acceptable but definitely could be better. Prior to digiKam, I was primarily using Photo Mechanic (no facial recognition) but it’s performance is slower than digiKam.

I originally felt PhotoStructure could benefit from facial recognition but after using multiple products for a while, I think facial recognition (and frankly, object recognition) may be best left for other products to do.

This thread is timely for me as I just recently switched my Photostructure install over to manual organization after I began using Digikam. This combination works really well for me as I enjoy using Digikam’s facial recognition features as well as its organization and light table.

My workflow now is greatly simplified:

  1. Plug in camera and import photos to Digikam
  2. Process photos, deleting some using Digikam’s light table feature
  3. Periodically run Digikam’s facial recognition process

Doing this Photostructure picks everything up. Now, I’m just looking forward to Photostructure’s favoriting and deleting and then I’ll have multiple interfaces to not only read from my photo library but also to write to it.

1 Like