After you start labeling faces, ACDSee’s facial recognition will try to find similar faces which have not been named and automatically suggest or assign names based on their similarity. Faces which are similar enough to be assigned a name automatically will then be used to find more similar faces. If ACDSee assigns an incorrect name, this error might lead to more errors unless the name is corrected. This is more likely when face detection scans a new person and there are no manual names for that person yet. Please note that suggested names are not used to find other faces, so incorrect suggestions cannot cause this behavior.
Finding and correcting every incorrect name manually can be difficult, but you can use the Rerun Recognition command in Tools | Options | Face Detection to automate most of the process.
When you choose to rerun recognition, all suggestions and automatically assigned names in the database are removed, leaving only the names you labelled or corrected yourself, suggestions you confirmed, and names imported from Lightroom or Picasa. Then, facial recognition runs again on all unnamed faces.
The most effective time to rerun recognition is after you make several corrections to names, since the new corrections will be used when rerunning recognition on other faces. You may still find a few incorrect names after rerunning recognition, but making a few more corrections and rerunning recognition again will help improve accuracy even more.
Rerun Recognition is also available through the Tools | Face Detection menu. The menu command will run only on currently-selected files. This can be a faster way to rerun recognition on a subset of your database, but any incorrect names outside of your selection could still affect recognition, so the command in Tools | Options | Face Detection is generally more effective. The menu command could be useful if you are confident about the names in most of your database, but have just added a new set of pictures that require some correction.
Modified on: Thu, 15 Apr, 2021 at 2:48 PM