Resetting and Retraining Face Recognition Database

Help
1

I have been using DigiKam for a long time and especially the face recognition feature, which has improved with each release. According to the press, it now achieves a recognition rate of 97%. I am currently using version 8.3.0 on Linux.

What is my concern?

There are “faces” that are reliably recognized and assigned to a person, which I can then confirm all at once. However, there are also “faces” that are incorrectly assigned to a person. I have the impression that for some individuals there is a good basis provided by the “Deep Learning Engine,” while for others, this is not the case.

I would like to reset the database (I have read that this can be done by deleting the “recognition.db”) and then retrain it based on 500 people with 10-200 images each.

Is this possible? If so, how do I proceed?

Thank you in advance for your help!

2

If you have written the metadata to the files, i.e. the face regions with name, then you should be able to delete the recognition.db and then retrain the model. If you don’t write the metadata to the files, then, from what I know, deleting the recognition.db will also erase all of your face tags assigned to the photos. I write new metadata to files, as this ensures the faces and other tags are always as close to the respective files as possible.