After uploading a few pictures last week, I spotted the ‘Add Name Tags’ link on my Picasa page. I knew this had to do with mapping faces to people, but what I had wrongly assumed was the simplicity. It took me less than a couple of hours to map about 2500 faces. I thought of sharing the accuracy of the tool & conditions that baffled it.

Unlike Orkut, where you have to mark faces yourself, Picasa automatically extracts faces. It took less than 5 minutes to extract some 5500 faces from about 3800 pictures. Another differentiator in the approach is that instead of mapping many faces on a photo (like Facebook, Flickr, etc), faces from many photos are brought together to be mapped to a person. This is what really saves time & effort and retains interest. Statistically speaking, atleast 60% of all my photos (about 1800 Nos) feature one from my family of four. So in the ideal case, I will finish mapping 1800 photos in 4 clicks instead of 3600 clicks (assuming 2 faces per photo). That makes Google what it is: not just an applications company, but a technology company.

I don’t know if people tagged by me will be suggested a priori to others in their photos.

The tool started off with my photos, possible because the count was the highest. I showed me about 12 full sets (x16) of my pictures. A couple of photos in the first set were quite old – about 12 years back. Soon, it started suggesting my name for all my photos.

The next set had pictures about 17 years old, from my thread ceremony; my head was bald then (like now).

This was the best: it could make me out in a black & white bald photo, and when I had a mustache, and a beard & a french beard. Awesome!

One of dad’s photo was also from the yester-years!

Kashish’s looks have hardly changed, but Picasa didn’t miss!

This is Chibu after Holi – wonder how Picasa got this?

This is my friend Vidyaranya ‘Vishal’ Rampurkar with part of his face covered. Picasa picked up his name based on previous training.

This was a bit crazy: it had grouped all my pictures but suggested my uncle instead of me. This should have been quite obvious given the tool’s performance so far.

This time it picked up all Chibu’s photos but included one of Isha’s. I really don’t think they look alike in any way. The suggestion was, however, correct.

This was quite expected, but we had managed to confuse the tool. Everyone says that me & Chibu looked very similar just before we started teenage. Picasa is not to blame here! 🙂

Now, the other way round! Chibu’s pictures were suggested as mine and the group did include one of my photos.

This time it got confused with my father and his brother, Rajesh Gupte.

Hemant looks like GBN Babu? Noooooooooo! (Dakshet style) Family tak theek tha; now the tool starts mapping people to their bosses… Lol…

Praveen surely mimics Manjunath sir well! I wonder if its that intelligent 😉

There was a single sardarji in all of my photos, and Google failed to recognize. I had already mapped Jasmeet ‘Jassi’ Singh a couple of times before.

At this moment, I got bored using the tool. The grouping & suggestion both went hay-wire. Some problem with Nikhil’s face though; it does not recognize him even today 🙁

Can this be used to verify if kids belong to a certain parents? 😉