Face tagging

Hi. I’ve just installed the docker version of Photostructure. I tried out the Lite version, liked it and am now trying out the Plus version under the free trial. I’ve resynced my library but don’t see any of the face tagging functionality. Do I need to rebuild the library to activate this?

What face tagging functionality are you referring to?

To be clear, Photostructure does not have a way to add tags currently. It can read and display tags that you’ve applied, but you cannot add new tags.

Adding tags, and face tagging specifically, are outstanding feature requests. You can read more about them in the following posts:

edit: Photostructure WILL pick up new tags that you add. So, if you add a tag in another program, Photostructure will pick that up on the next sync.

Ah, OK. Didn’t realise that. I’ll wait on v1.1. Thanks for the help. :slight_smile:

Do you mean if program writes tags directly to image file? Matthew once said about yaml support, which can be located in some other place, like PhotoPrism do, but I did not heard it was implemented?

Photostructure will read tags embedded in the image file, or in a sidecar file, and it’s been that way since the beginning. @mrm will have to comment on anything beyond that!

I mean I really do not know any good examples I can edit metadata and in will be seen in PhotoStructure. I use Shotwell and it can actually edit original files, like write tags to it, but that is so bad idea, the feature itself is behind a switch in settings turned off be default. Another one I use is PhotoPrism, AFAIK everything you edit are stored in yaml (which are stored in separate folder from image and json) or database (I am not sure actually), json itself is never edited 100%. And manual json editing is just beyond me and the whole idea to keep things simple.

I don’t know if I’d say it’s a bad idea… I use TagThatPhoto for face tagging, which writes it’s tags directly to the original files. The risk in writing to original files is that if the write isn’t successful, the image might become corrupted. This risk is completely mitigated if you have backups of your photos before you start writing tags… and you SHOULD have backups of your photos no matter what.

The other software I’ve heard of that people use for tagging is Digikam. I’ve never used it myself, so I can’t speak to how good it is.

Yeah… I believe anything written in image actually corrupts it. I mean sizes and crc will be different and duplicates hell is so close than :grin: that’s why photo library software do not do it, all tags are stored in sidecars. The problem is sidecars files could be different, Google uses json, yaml is more common format, some software store it in databases, all those methods are OK, but each app uses own method and that is a problem.

They don’t do it because they don’t want users blaming them IF an image gets corrupted and they don’t have backups.

Obviously you should do what you feel comfortable with. I’m just pointing out that there are LOTS of people that are comfortable using software that writes embedded tags that don’t experience issues, and they aren’t crazy… they just know the risks and mitigate them.

Sorry, just seeing this thread now.

PhotoStructure writes orientation and ratings metadata currently—and you’re correct, I was concerned about rewriting metadata to files, especially if they weren’t standards-compliant already (many files throw warnings within exiftool for noncompliance or invalid encoding), but for things like orientation, no applications read from sidecars for that field, so when I was writing orientation to a sidecar, and nothing else picked up that change, it looked like there was a bug in PhotoStructure.

There isn’t a standard yaml sidecar that I’m aware of. You’d want to avoid yaml in any event, because it’s such a difficult and complicated file format to parse correctly.

There are more than five advanced settings within PhotoStructure to control metadata extracting and persistence: you can choose to either wholly use sidecars (and pick from a standard, like XMP, EXIF, or MIE), wholly avoid sidecars, or pick and choose which tags go where, even between videos and images.

A future build of PhotoStructure will actually allow for both file SHA and image SHA, the latter of which is robust against metadata changes

Once I started structuring my library I tried lots of software and tools, I had to deal with images without date, or without date in exif, but a date in sidecar. I started again and again and again. I knew so much shit …jhead, exiftool, podman, SQLite, MariaDB. I aced rsync, rclone and syncthing. Luckyly duplicates problems were not a big mess, but some dup problems I created myself :smile:. There is no perfect library software, I now use simple Shotwell, powerful PhotoPrism and beautiful PhotoStructure. While they both use same Originals folder, they do not easily share metadata between themselves, and also all of them has own problems and own benefits.

speaking about digiKam… it’s configurable. You can have it write the metadata in XMP sidecars, in the file, or both! And photostructure can deal with all scenarios just fine.