This is painful. You haven't even bothered to use a label on the radio buttons, so my click area is larger.
But even had you done that, it would only have been mildly less painful.
Use every inch of screen real-estate available on well defined click areas, so people can rapid fire answer the questions without it becoming a mouse accuracy test.
Better yet, Ask a single question at a time, and at every mouse click, bring up the next question (again using all the screen area available, so an over-caffeinated person on a trampoline can still hit their click points easily). After the last question is clicked, automatically bring up the next picture, and start the procedure again.
Ask only 1 question about the picture with answers with labels and keyboard shortcuts. Pictures would get answered partially, but could be in the queue until fully answered.
I think there is a huge opportunity in chatrroulette if we can just solve the "penis problem".
I've used chatroulette a few times over the past week and if you find someone "normal" to talk to it can be a lot of fun. That said, you spend most of your time "nexting" perverts.
I've been toying with eye-tracking using computer vision for assisting individuals who find themselves paralyzed after a car accident, stroke etc. But I'm not (yet) a CV expert, so I have a long road ahead of me. Part of my desire in this field is to have the computer boot and automatically detect the face and specific facial features from the webcam, with no calibration.
There is a lot of existing work in this area and it seems to me if we can detect a mouth on a face in a video stream we could use the same methods to detect someone jerking off.
I think the hardest part in the process would be finding an army of labelers that would put up with labeling the disturbing imagery. (Can you even submit this kind of material to mechanical turk? I'm not sure.)
Sure I'd have no problem with that. There is a vote convergence system in the backend so there will be a bunch of raw votes on each image that I could easily export if someone wanted them.
Do you have plans for the data? Are you going to give others access to it?
I pitched in a little, until I couldn't take it any more :) A way of hiding/blocking an image while you answer might be helpful for this sort of thing.
Maybe I'm too obtuse, but the instructions suggested to me not that you wanted us to categorize the image, but, since the image I got was of an empty room with a sofa, it was a placeholder and you were going to bring back an image that matches the "search" criteria...
Where are you getting this image data from, out of curiosity?
Also, any plans to publish a live report? I'd be more inclined to keep helping out if there were some interesting graphs I could look at; the promise of a future report is less enticing :)
But even had you done that, it would only have been mildly less painful.
Use every inch of screen real-estate available on well defined click areas, so people can rapid fire answer the questions without it becoming a mouse accuracy test.
Better yet, Ask a single question at a time, and at every mouse click, bring up the next question (again using all the screen area available, so an over-caffeinated person on a trampoline can still hit their click points easily). After the last question is clicked, automatically bring up the next picture, and start the procedure again.