I assume each time an additional tagging is made of a person in the data base, the data base is updated and the person recognition will improve. My question is about manual face marking and tagging for cases not detected automaticly. In some cases the tagging is of a the person which I want to be able to find in a search, but I do not want LR to "learn" from this example as he is wearing a mask / viewed from the back / standing on his head. What should I do? Does LR "learn" and update the data base also from manual face marking? If not, there is no problem. If it does, do these "bad" manually added examples harm the recognition or not.