Tag Images With Your Mind

blee37 writes "Researchers at Microsoft have invented a system for tagging images by reading brain scans from an electroencephalograph (EEG). Tagging images is an important task because many images on the web are unlabeled and have no semantic information. This new method allows an appropriate tag to be generated by an AI algorithm interpreting the EEG scan of a person's brain while they view an image. The person need only view the image for as little as 500 ms. Other current methods for generating tags include flat out paying people to do it manually, putting the task on Amazon Mechanical Turk, or using Google Image Labeler."

Comments

Looks Good on Paper...

[Anonymous Coward]

Honestly this is nice, but seriously if my mind was tagging my images there would be something like the following list of tags
idhitit, awesome,ohgod, thehelliswrongwithme, whydidisavethisagain, ohthatswhy, shit, wallpaper, and photoshop
and that's just keeping it within PG-13.

Re:Looks Good on Paper...

[dominious]

if my mind was tagging my images there would be something like:

porn

Re:Looks Good on Paper...

[Anonymous Coward]

So you frequent 4chans /b/, eh?

Re:I mentally tagged this as MenWhoStateAtGoats

[camperslo]

My Steve Ballmer pictures keep coming up as "best used as cat food"

Cool!

[Rik Sweeney]

Just don't let Rorschach tag any images of ink blot tests.

Boobs boobs boobs

[erroneus]

Microsoft, be warned. Some people have a limited scope in terms of what they are thinking about at any given time.

I was tagging that image, honest!

[noidentity]

Honey, I wasn't looking at her breasts; I was just tagging the image using Microsoft's new mind tagging, honest!

Fun and Easy to Use

[smitty777]

This is typical of MS -thinking that something like this would be easy for the average user. FTA: "However, the mind reading approach has the advantage that it does not require any work at all from the user."

So, in order to use this sytem, we should all strap on EEG caps while we're surfing the web. Sounds real practical to me - I used to work in an EEG lab, and I can tell you that those caps are pretty uncomfortable to wear. After they put them on, you stick these little needles into the leads and squirt conductive goop on your scalp. It takes a few cycles to rinse that stuff out too.

Way to go MS for making productivity so much easier.

Re:Fun and Easy to Use

[DynaSoar]

I used to work in an EEG lab, and I can tell you that those caps are pretty uncomfortable to wear. After they put them on, you stick these little needles into the leads and squirt conductive goop on your scalp. It takes a few cycles to rinse that stuff out too.

Smitty, we've come a long way from those caps. There are now "caps" that are essentially nets of elastic cord with plastic cups containing pieces of sponge in them, the electrodes embedded in the sponge. Dip it in mild salt water for conduction, shake it out so there's no drips running together bridging the electrode sites, and pull it on. I could get good signal on 128 channels in less than 10 minutes from the time they walked in to data collection start.

There is also a European company selling a similar get up, but the preamps are built into the cups on the net, making impedance matching irrelevant and signal balancing automatic on the fly. These are so stable that they can be used ambulatory.

And nobody ever has to get goop or glue stuck on/into them any more.

Re:Oh Microsoft...

[smitty777]

Actually, the variance isn't the problem. That comes out statistically in the wash - you can see that with a large enough N, patterns emerge across the different stimuli types, which allows them to do the tagging. The real problem is interpreting the complex interactions between the different regions of the brain. However, that doesn't really matter for an experiment like this, as the patterns don't actually need to be interpreted, just recognized by the algorithm. It's a similar concept to the way the MMPI psychological test was developed, in some ways.

It's funny, the MS team doesn't seem to even recognize this. FTA: "and scientists’ currently weak ability to interpret what brain scans mean.

Oh that what the internet needs more of!

[Interoperable]

Tags.

no semantic information? Ahem *cough*

[davidwr]

unlabeled and have no computer-readable semantic information.

There, fixed that for you.

Seriously, the old saying "an image is worth 1,000 words" implies that images frequently have semantic information, at least in the sense that anything on paper can have semantic information. It's just that computers can't parse it, catalog it, search on it, etc. Not well, anyways, not yet. They are getting there though.

3-class

[Metasquares]

Useful, but real-world tagging is much more specific than "person", "animal", or "inanimate". The number of classes required in the classification task is thus far greater and one would expect the accuracy to be proportionally lower. OTOH, it could be a great preprocessing step for further manual analysis, or a step in a hierarchical clustering algorithm. Or maybe 3 classes suffice for certain specific situations.

Focus

[mumb0.jumb0]

What happens when I'm tagging a photo but listening to music at the same time?

Or I run the photo tagging software in a small window and watch a movie (or some porn) instead?

So they can create tags from brain waves, but there's no way to tell what a user is actually focussing on.