Casual Fridays: More mystery photos!

Remember these photos from earlier this week?

i-36491118462baccd4d2ee14464482f77-tinypix.gif

Readers were intrigued with the idea that we can tell which photo is a face, despite the fact that the photos are just 12 by 14 pixels! That brings the question: can we identify faces with even *less* information? This week's study may help answer that question. You'll be presented with eight images of different resolutions. Some will be faces, and some won't be. Can you tell the difference?

Click here to participate

As usual, the study is brief, with just 8 quick questions, so it should only take a minute of your time. You have until 11:59 p.m. Eastern time on Wednesday, February 28 to participate -- or until we have 333 responses, whichever comes first. Then don't forget to come back next Friday for our analysis of the results!

More like this

These pictures reminded me of my visit last weekend to the Star Wars exhibit at the California Science Center in L.A. They have several interactive experiments that show what someone with a retinal implant might see and how robots can recognize faces by looking for the eye/nose pattern in images like these. You can try to fool the robot by having multiple people in its field of vision, or by turning your profile, but it's surprisingly good at recognizing when something human is standing in front of it.

The rest of the exhibit is fun too. Well worth a visit if it comes to a museum near you.

I don't know if I got any right, but it sure seemed easier when I was leaning way back with my glasses off. Probably would have been even easier than that if they were in the original size.

How are you shrinking the images? I ask because just today I was reading the special section on "Modeling the mind" in Science from October 6, and seeing these images now made me think of this paragraph, in the article on Eero Simoncelli's research (on page 79):

Last year, Simoncelli and his colleagues reported building an image-compression tool based on his nonlinear model of cortical neurons. Simoncelli reasoned that if the brain's visual cortex is optimally efficient at processing images, it should also do a superior job of compressing them. What's more, any distortions introduced by his compression process should be tolerable. "If the cortical representation is like what's in our brains, we won't notice the difference," he says. Indeed, the new compression technique's performance far outstripped that of the JPEG standard.

Wouldn't it be nice to use that and see if our face recognition perfomance improves?

I once read that a 4x4 pixel representation of Abraham Lincoln was recognizable by some large percentage of Americans. I've been trying to find reference to that for some years now.

The mind is a strange thing ... I could have sworn that 6th image was of Napoleon (I am waiting for the answers to see how completely off I was). The lower resolution ones were very difficult and I have no idea whether I guess correctly on any of them.

By K Bescherer (not verified) on 23 Feb 2007 #permalink

Wow--

That study filled up fast. Too bad we didn't do this one last time, when we had plenty of survey monkey credits. I may try to set up a polling option to allow more people to participate.

argh missed out again