I have being using PSU since 2015 (v3 back then, and now v4), after a long due diligence looking for the best DAM for my needs. So far I am very satisfied with the choice: PSU is not perfect (no piece of software is) but it's still the best for me. Among the missing features is facial recognition, in my opinion a must-have feature for DAM software in the future (some of them have it now already).
Last week I come to know of a recently (April 2019) launched tagging software sporting facial recognition and I tested it over the weekend. Its name is "Tag That Photo" and it does just that: it detects the faces in my pictures (PSU does this as well) and it recognizes whose face it is. So that once I tag one face it automatically tags all the others belonging to the same person. Then it writes the tags and the areas inside the image file in the proper XMP fields (and it keeps its own database of course). So when I "read metadata from file" inside PSU I have my picture labelled and tagged within PSU.

Since PSU is still missing the feature (next major release maybe?) for the time being I will be using TTP to complement it. Apparently a single piece of software can't handle all the functions in the image processing chain well enough: now I have PSU for DAM, DXO for raw processing and TTP for tagging...horses for courses I guess...