+1 news about this feature??
I’ve some details in here about my setup:
+1 . Lack of facial recognition unfortunately made med cancel my Plex Pass subscription.
+1 I bought my Premium pass banking on this becoming a feature. Please don’t make regret my decision.
It would also be nice to have the option to use external recognition engines. E.g. Azure Media Services or so.
Plex seems to have committed itself to imagga for auto-tagging. imagga advertises itself for facial recognition: Imagga - Image & Face Recognition API
Is there hope that Plex will leverage imagga for this face recognition feature? I have searched the imagga website, and face recognition only seems to be available as an on-premise service, which isn’t the service that Plex uses. via Plex>Library>Edit>Options: photo thumbnails will be temporarily uploaded to a third party for analysis. Learn more
My intention is not to oversimplify the problem, but to advance the conversation. As I mention above, imagga does not seem to have face-recognition-as-a-Service (FRaaS?); I’d happily be proven wrong here.
I was researching other FRaaS alternatives and found Betaface API and LambdaLabs API. As a slight tangent: another interesting option for Plex may be for each customer to host their own using a OpenCV installation and processing. The benefit with the latter is that OpenCV seems video capable.
Lets explore the betafaceapi demo because it was downright easy and seems to be the approach that Plex takes with imagga; first uploading customer thumbnails to a service and then receiving results.
To follow along, please reference the following three images that I posted on imgur: 1 person 6 perspectives, 4 people 2 perspectives each, 54 people 1 perspective each.
Repeat for each of the three images: Copy the URL, paste it into the text window in the betafaceapi demo, and click on “start detection”.
** I’m using three files with multiple faces for simplicity. you are more than welcome to upload 68 individual portraits for your own testing
Once you have uploaded all three:
-
Click on the first file, then a face, and lastly click on compare faces. You will see 80%+ confidence in matching his 6 perspectives, and lower percentages with other faces (this is just using one face and not trained on 3 or more as suggested).
-
Click on the second file, then the face in the upper left, and lastly click on compare faces… you will see a similar result. 90%+ confidence on her smiling face, and lower percentages with other faces.
-
Click on the third file, then any of the 54 undiversified faces, and lastly click on compare faces. Even though they are all similar, the face recognition shows 80% or less confidence on any of the other faces in the 3 files.
One other important feature: once you identify a face, you can use the “set person” function to associate the face to a name. The more images you identify with that name, the more you will train the FRaaS to identify that person in your images.
How can it be incorporated into Plex? it seems that it can be independent from imagga. perhaps: Plex>Library>Edit>Options and include a checkmark to “Automatically add Facial Recognition to my photos” with a similar disclaimer: photo thumbnails will be temporarily uploaded to a third party for analysis. After FRaaS computation, there could be a process outlined above to “set person” as a new or existing named tag. This FRaaS can be separate from imagga tagging service, but feed into a common tag library for each image that I assume resides on the customer’s host computer. You will then be able to search through a common source of all your tags.
Although imagga gives me the ability to search for a smiling woman in a corn field, with a FRaaS such as betaface incorporated into Plex, I will finally know that the smiling woman is in fact: Mary, and not Margaret or Molly. And THAT is what Plex is painfully missing in its photos library.
Again, here is the Betaface API
I hope this helps.
+1 for facial recognition/auto tagging of person which can be defined
+1 for facial recognition. This is really the one feature which would allow me to really use plex. The current autotagging has been a bit of disappointment, as the generated tags are not really useful for much at all: #happy #child #family is basically put on all pictures.
It has been stated in the thread that “this is definitely something plex is looking into”… is there any plan or timeline when this might be released? Is it even being worked on?
Thanks for this post, hope this can help to improve the avancement of the facial tag, waiting the implementation into Plex.
I Don’t know how we can help as user, but if help is needed, I can use my time to avance on this topics (IT Development guy). In this new GDPR World, this specific feature will have some difficulty, unfortunately.
I would like to have my local nas do facial recognition / tagging / … The device can take long and may be inefficient, but it is idling the whole day. The facial recognition is my killer feature for using google photos and I would love to take home my pictures as well.
+1 for locally hosted facial recognition & tagging.
It would be great if it could import existing tags (mine are mostly facially tagged by MS photo gallery) & would be even better if you also had a utility that ideally could be run on the Plex server host separately from the Plex service itself (as it would need write access to the media whereas Plex would often only be granted read) that could read the tags maintained within Plex & add any new tags into the relevant media files metadata so that they can be shared with other applications.
Also I’m not sure if something is wrong with my instance or if this is a bug, but although I have many tagged photos in my library, I’m unable to search by tag via my IOS plex app. Once I’ve opened a photo, I can see the tags & jump to similarly tagged photos, but I can’t apply a tag based filter. Regardless of what search string I type, there are zero results.
Heh. Facial recognition for people with terabytes of media. Just what the lawyers want.
+1 to facial recognition request. This is the most important feature to me and unfortunately without it is forcing me to look at alterative options
I’ve now got plex app dumping all our photos from our phones into our ubuntu server nas folder.
Are there any docker apps or similar that will tag locally hosted images using facial recognition?
It looks like Plex’s Partner Imagga now supports Facial recognition: https://imagga.com/solutions/face-recognition-and-detection
Plex devs, is this something that can not be built in?
Bump. Just making sure the topic doesn’t close. Is this feature planned?
The single ‘employee’ that was in this thread is no longer an employee and his last post was April 2017.
That pretty much says it all.
Please add this! And please add for places that you can draw an area on the map from which you want to choose pictures for a dia show e.g…
Definitely need us some facial recognition. It shouldn’t be to hard to implement. Autotagging looks tight but I would like to search for my fam a hell of a lot more than some fruit.