Has anyone worked on adding in support for AWS Rekognition? seems like an
excellent way to obtain AI based tags, metadata etc and then for digiKam to manage the photo library. -- Sent from: http://digikam.1695700.n4.nabble.com/digikam-users-f1735189.html |
Hi, Do you mean to use Amazon face recognition engine ? To use cloud web service to recognize peoples ? You are serious ? Face tags (people tags) are private and we never share these information with a cloud web service. Best Gilles Caulier Le dim. 20 oct. 2019 à 22:08, tb582 <[hidden email]> a écrit : Has anyone worked on adding in support for AWS Rekognition? seems like an |
I agree. Some kind of AI to recognize elements in a picture would be cool,
but I wouldn't like my pictures to be uploaded to some server, specially not Amazon/Google/Facebook/Apple/etc. Maaaybe if it was some non-profit server, and there was a guarantee that they wouldn't keep the data, I could think about it. Like the current use of Openstreetmaps for reverse geolocation. -- Sent from: http://digikam.1695700.n4.nabble.com/digikam-users-f1735189.html |
Yes - AWS allows you to upload local byte streamed encrypted files for
analysis- no storage on their servers, analysis is done, image is purged and you get all the good data. https://docs.aws.amazon.com/rekognition/latest/dg/images-bytes.html -- Sent from: http://digikam.1695700.n4.nabble.com/digikam-users-f1735189.html |
In reply to this post by woenx
In all cases, with next 7.0.0 release, the lead advancement will be to drop all current face recognition algorithms in digiKam to be replaced with new OpenCV based Deep-learning algorithm which give excellent results here... Take a look to this video : Be patient. Best Gilles Caulier Le dim. 20 oct. 2019 à 22:37, woenx <[hidden email]> a écrit : I agree. Some kind of AI to recognize elements in a picture would be cool, |
Interesting but that only helps with faces, what about object, text etc as
AWS Recognition ? -- Sent from: http://digikam.1695700.n4.nabble.com/digikam-users-f1735189.html |
In reply to this post by tb582
On dimanche 20 octobre 2019 22:55:55 CEST tb582 wrote:
> Yes - AWS allows you to upload local byte streamed encrypted files for > analysis- no storage on their servers, analysis is done, image is purged and > you get all the good data. Who guarantees that purging? And even if they purge the images, what about the results of the analysis? I'd expect those results to be a lot more interesting for Amazon, especially since the results are linked to a user account. And they carefully avoid mentioning analysis results when they talk about "content processed". Oh, and it's a paying service, requiring registration. Not necessarily a problem if you think it's useful, but would have been nice to see it mentioned... For me, no thanks. (I don't need face recognition, and I doubt the service is precise enough for my wants). Remco |
In reply to this post by tb582
> Yes - AWS allows you to upload local byte streamed encrypted files for
> analysis- no storage on their servers, analysis is done, image is purged and > you get all the good data. You believe that? Why should they provide such a service if no interesting data is generated? The latest allegations about the cloud surveillance cameras did you notice? Maik |
In reply to this post by tb582
Le dim. 20 oct. 2019 à 23:28, tb582 <[hidden email]> a écrit : Interesting but that only helps with faces, what about object, text etc as OpenCV has already all the infrastructure to detect and recognize something like object, monument, plant, animals, etc... When i will switch to 7.0.0 code, i plan to make the detection and recognition as plugin and make it extendable. I'm not yet sure, i must go in-deed in source code before... Best Gilles Caulier |
Gilles Caulier-4 wrote
> Le dim. 20 oct. 2019 à 23:28, tb582 < > tb582@ > > a écrit : > >> Interesting but that only helps with faces, what about object, text etc >> as >> AWS Recognition ? >> > > OpenCV has already all the infrastructure to detect and recognize > something > like object, monument, plant, animals, etc... > > When i will switch to 7.0.0 code, i plan to make the detection and > recognition as plugin and make it extendable. I'm not yet sure, i must go > in-deed in source code before... > > Best > > Gilles Caulier Thanks - is there a ballpark timeline for 7.0? Weeks? Months? -- Sent from: http://digikam.1695700.n4.nabble.com/digikam-users-f1735189.html |
Not yet. We need to finalize 6.4.0, probably in November, and switch master to 7.0.0. This must be done before to start next Google Summer of Code planed in March 2020, so in December 2019... Code from 7.0.0 development branch already work and is sync with 6.4.0, but code need to be clean and old implementations removed. Best Gilles Caulier |
In reply to this post by tb582
The more people contribute code, the faster it'll come ;)
On October 21, 2019 9:59:34 AM PDT, tb582 <[hidden email]> wrote:
Gilles Caulier-4 wroteLe dim. 20 oct. 2019 à 23:28, tb582 <tb582@> a écrit :Interesting but that only helps with faces, what about object, text etc |
Free forum by Nabble | Edit this page |