Photo Tagging with Windows Live Photo Gallery

Microsoft’s Windows Live Photo Gallery beta was released June 2010 and I’ve found it quite addictive. It makes tagging people’s faces quite fun – almost a game. In the first eight days of using this new version I tagged 6310 people in my photos (245 unique people). That’s almost 800 people a day!

The addictive thing about WLPG is that when it is first run it starts scanning your images, looking for faces. It does a pretty good job of identifying human faces, although it occasionally makes some amusing mistakes, like this chain link fence which it thinks contains five faces:

It took WLPG about sixteen hours to process my 33,000 photos for faces. Then the magic started. I clicked on “Batch people tag” and it said “Are these John Smith?”, while showing me a half-dozen pictures that were indeed of John Smith. That seemed pretty cool so I clicked “Select all” and “Confirm” and it then showed me a dozen more pictures (of John Smith) and again asked if these were him.

Each time I confirmed its guesses (which were probably about 95% accurate) it seemed to gain a better idea of what John Smith really looked and it would find more pictures of him. Apparently I had previously tagged a dozen or so people in my photos – just enough to give it a starting point – and it was now using that information to help me tag photos at a previously unheard of rate.

I treated this like a game and tried to see how quickly I could plow through its suggestions. Whenever photos of one person temporarily ran dry I’d switch to somebody else, only to find that it had then found even more photos of the first person. When it finally slowed down I started hunting through important family events and five-star rated photos to start telling it about new people. Given my poor memory for faces and names this then turned into a research challenge. I used my family tree, e-mail search, yearbooks, linkedin, and many other resources to track down the names of people in photos that were up to twenty years old.

Then I found that some of my casual acquaintances were tagged more heavily than my favorite relatives, so I had to hunt down photos containing my favorites so that they didn’t appear to be losing the popularity contest. And when one of my daughters complained that there weren’t enough photos tagged with her I had to fix that; until the other daughter complained.

And that’s how eight days of compulsive/obsessive/productive photo tagging went. At the end of it WLPG had (mostly) stopped giving me new useful suggestions. I’d tagged everyone in all 2,200 five-star photos, and I’d added 245 unique people to my photo database. All of this information is conveniently stored in the images (rather than just in the photo database) so that when I sync the photos to other computers or back them up I am backing up all of the names.

A graph of the distribution of photo tags by people shows some interesting trends. I tagged as many photos of my daughters as of the 235 lowest ranked people combined, which seems appropriate. My spouse’s poor position relative to me, however, suggests I may be an egomaniac, or at least potentially in trouble.

Now I can easily do searches for things like five-star-photos containing both daughters, or with a bit of geotagging, all photos containing all of my siblings that were taken in my home town in 1998.

Cool.

Good fun. Recommended. Do it before you forget who you invited to that barbecue ten years ago.

About brucedawson

I'm a programmer, working for Google, focusing on optimization and reliability. Nothing's more fun than making code run 10x as fast. Unless it's eliminating large numbers of bugs. I also unicycle. And play (ice) hockey. And sled hockey. And juggle. And worry about whether this blog should have been called randomutf-8. 2010s in review tells more: https://twitter.com/BruceDawson0xB/status/1212101533015298048
This entry was posted in Computers and Internet, WLPG. Bookmark the permalink.

3 Responses to Photo Tagging with Windows Live Photo Gallery

  1. Pingback: Basics of Digital Photo Organization | Random ASCII

  2. Jennifer says:

    I’ve just recently discovered batch photo tagging and I also find it addicting. I’m obsessed with working on it every day until every last photo is tagged. I find it extremely accurate but have had a few humorous suggestions. For example, I have one particular table lamp that keeps coming up as a face. However, I had a picture of my son standing in front of the Christmas tree, and a very small picture of him from an ornament hanging on the tree showed up as a suggestion. I couldn’t believe it picked up such a small photo in the background and accurately suggested the correct person.
    I would like your advice on what you did with either people you didn’t know, like someone standing in the background at the zoo or other public place. Another situation is a school concert or other such event where I took a picture of my child on stage and dozens of other kids standing around him are popping up as suggestions and I don’t really care to tag them.
    Thanks for the article you posted!

    • brucedawson says:

      The misidentifications are amusing, but sometimes revealing. WLPG thinks that my brother looks like me, and it’s not wrong.

      When the face identification notices faces of strangers I delete them. It is unfortunate that WLPG does not have an efficient way to do this — it is tedious when twenty faces are found in the background.

      If you delete all of the faces in a picture then WLPG may scan it again sometime — if you move it such that it thinks that it is a new photo. So deleting all of the faces may not be worth it if you can’t tag a single face, ’cause you may need to do it again.

      Perhaps this behavior can be used. Just tag one face, then rename the photo (using Windows Explorer, not WLPG) and when WLPG notices the ‘new’ picture it shouldn’t do any face scanning, so it should just have the one face.

      I’m glad you’re enjoying.

Leave a comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.