![]() I guess is good to do it from time to time when new people is added to your collection and/or when you have matched a lot of faces and people, specially if you have fixed many mismatched faces. What I don't know is when you need to re-train the database. IMPORTANT: On the settings sugbtab, select the YOLO v3 detection model. I guess this 'trains' the database using the faces already detected and assigned.įinally, when you hace the database trained, you can go to the People tab again, select "Recognize faces" and let Digikam match detected faces with people. Go to Tools > Maintenance and select the update faces. But first, and this seems to be key, you have to instruct the app the train the faces recognition database. Once you have a minimum set of faces, you can use the recognize feature. Once you have faces detected assign each faces with a person. ![]() I would start with a limited set of albums. On the workflow subtab (at the bottom), select "Detect faces". How to detect and assign faces? You can do it manually (editing a picture and adding a face), or you can go to the People tab and use the detect faces feature. If not, you will need to do some more work. So, you need some faces already in place (this means detected and assigned to someone). The thumbnails have Tool Tips which will pop up if you allow the mouse to dwell over a thumbnail for a moment. Select the Album View page and look for the Icon-View Options section. I'm not totally sure how it works, but AFAK the feature needs some training. You can control what information is displayed from the Settings Configure digiKam. ![]() On first start up, application needs to download some external data for the face recognition. First one is to detect faces(detect parts of the picture that are faces), and the second one is to recognize the faces (assign a detected face to someone). ![]() Keep in mind that this woprks in two phases. It seems Digikam 7.2 improved this feature. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |