I’ve been playing around with iPhoto on and off ever since Complexli bought the Mac mini for me.
In particular, I’m interested in face recognition. It would be so nice if all your photos could be tagged correctly to show who’s in them.
However, I’m less than impressed with iPhoto’s performance in this area.
One thing is that the face recognition on its own is quite bad. I appreciate it’s a hard thing to do well. It appears to me, though, that it’s doing a worse job recognising the blond members of the family, such as Léon, than spotting the darker ones, such as Anna. I read somewhere that it’s been developed by a Japanese company, which might explain this.
However, I have two major complaints.
Firstly, if you upload a lot of photos in one go (I tried about 15,000), it will keep processing faces in the background forever. It’s now been running almost constantly for over a week, so I think it might have entered an infinite loop somewhere. This means that it will almost never make a suggestion, and it keeps one processor 100% busy constantly, which is not great for system performance.
Secondly, it seems not to use any other information when making guesses.
For instance, if ten photos have been taken within half an hour, and the nine of them contain only Léon, wouldn’t it be reasonable to suggest that the tenth also contains him? iPhoto doesn’t think so: It will happily suggest it’s Léon on one photo, Anna on the next one, then Gordon, then Charlotte, and so on.
Also, it doesn’t seem to take dates into consideration. If there’s a baby in a photo, it’s a good guess that it might be Marcel if the photo is from 1997 or 1998, Charlotte if it’s 2000, Léon if it’s 2005 or 2006, and Anna if it’s 2007 or 2008. But iPhoto will happily make suggestions that are not reasonable given that all photos have a date stamp.
It’s so frustrating to know that it could easily be improved, but it’s closed source, so I can’t do anything.
I prefer open source!