Wow! What a difference the most recent upgrade to Picasa has made. Google has once again upgraded their FREE photo management and editing program with two huge new features - geotagging from within Picasa (which I haven't tried yet) and Face Recognition (aka Name Tags).
I started working with the name tagging feature last Sunday morning. Basically all you do once you upgrade to 3.5, is launch Picasa
and let it do its magic and it will find all the faces of all people (dogs, too) in all your photos.
For most people with photo collections of a few thousand photos, the process to find all the faces can go relatively quickly - maybe a few hours. As I've mentioned in other blog posts, I have in excess of 100,000 photos dating back to 1996 and a lot of those photos have people in them. For my system (duo core processor, 4GB RAM running Windows Vista), my system has been on and running since Sunday morning and the scan is currently (Weds at 6:50 p.m.) at 13% completed. At this rate, I'm estimating it'll be well past Thanksgiving by the time it's done.
So, why am I doing it? For me, it represents a huge leap forward in managing photos. If I want to find all the pictures of my son or daughter now, it's literally impossible to do. If I want to find my son at a soccer game, I might be able to find them, but over time, it becomes more and more difficult do so. Now along comes "name tagging" and now the task is much, much simpler and much, much quicker.
Face recognition isn't new, but from what I've read, Google has made the process the simplest out there. In the past, you could have gone to each photo and added keywords with names, but the process was very laborious and difficult. Google has made the process very, very easy and (as usual) intuitive.
I don't know how it does it, but it goes to each image, finds faces and crops the photo (virtually, of course), so all you see is the person's head. It then tries to cluster similar looking people together and if you identify one of the ones in the cluster with a name, it'll identify the others in the cluster the same way. It's really pretty slick how it works.
If Picasa identifies somebody in a photo you don't know or don't want to tag, you can click the X on their picture and it will ignore that person in the future.
Here's the problem though: when it scans your images, it looks for every person in every photo. So, if your subject is a large group of people, it grabs ALL the faces and asks you if you want to identify every one of them - even if they're in the background and their image is blurry, it still tries to get you to identify them. The photo on the left is an example of a shot that Picasa will try to grab every face from.
So, with my 100,000 photos, multiply that by some number (e.g. average # of faces per photo) and literally, there could be millions of faces to identify. Daunting task, to say the least.
One thing I haven't been able to find, are real instructions on how to best use this technology. There are some basic instructions, but no "best practices" to help people get better results in less time. So in a future blog post, I will jot down some ideas of what I think are Picasa name tagging "best practices." If you follow some of these concepts, it should save you some time and you should end up with better results.
So, in my next blog post, I'll list some of those best practices and some ideas on how Google could make their already awesome program even more of a killer app. If you're reading this and you haven't tried Picasa yet, go out there, download and install Picasa 3.5 and let the magic begin.
No comments:
Post a Comment