Support NeoGAF

[McFly]

http://www.ps3focus.com/archives/131

The booth of joint Cell developers, Sony, Toshiba and IBM featured a real-time “digital mirror” simulator powered by the cell processor. A model was seated in front of the simulator that was then able to display a 3D model of the person that matched the movements and facial expressions (!) of the person in real-time.

In order to create a 3D model of the face of the person sitting in front of the camera, the processor first mapped 500 coordinates. Details such as hair and makeup were then superimposed onto the 3D model. To keep up with changing facial expressions and positioning, all 500 coordinates were read 100 times per second. Different hair, makeup and tattoos were applied to the model in realtime.

The “digital mirror” demonstration is in principle an Eye Toy setup, where a camera is able to detect movement and track objects, albeit with a much higher resolution.

Sony also screened a video demonstrating that these techniques could be applied to simulate the full range of human movement.

Eye-Toy 2? Make it happen Sony!

Fredi

 

[Varian]

In order to create a 3D model of the face of the person sitting in front of the camera, the processor first mapped 500 coordinates. Details such as hair and makeup were then superimposed onto the 3D model. To keep up with changing facial expressions and positioning, all 500 coordinates were read 100 times per second.

Sounds pretty fine-tuned. A non-Japanese face would probably make it explode.

 

[miyuru]

Sounds pretty fine-tuned. A non-Japanese face would probably make it explode.

:lol

 

[Wollan]

Sounds pretty fine-tuned. A non-Japanese face would probably make it explode.

:lol :lol :lol

 

[Panajev2001a]

Sounds pretty fine-tuned. A non-Japanese face would probably make it explode.

Someone needs to meet some Japanese girls IRL they are even more expressive than many western ones.

 

[Wollan]

'Decode this, bitch.'

 

[sly]

'Decode this, bitch.'

OMG! MY EYES!!!!

 

[Varian]

Someone needs to meet some Japanese girls IRL they are even more expressive than many western ones.

Jesus Christ. The point is I don't think the hardware can read such precise data from different ethnic groups when their features differ so wildly.

 

[Twix]

sounds great.. Eye Toy 2 should be Sony’s real surprise

 

[Panajev2001a]

Jesus Christ. The point is I don't think the hardware can read such precise data from different ethnic groups when their features differ so wildly.

Ok, I took it differently and a bit more bitterly, but I guess it was the streams of lol's that helped too .

 

[empanada]

Sounds pretty fine-tuned. A non-Japanese face would probably make it explode.

This face will definitely make it explode.

 

[Kleegamefan]

'Decode this, bitch.'

:lol :lol :lol :lol

 

[Suikoguy]

I thought this was not premapped. That means the tech is already ready for reallife use. As someone mentioned in the other thread, this has many realworld uses, especially in trendy locations. Hairdressers could make the biggest use of this, but other places too.

This is the next-gen persay of human-use technology for everyday use.

 

[genjiZERO]

Jesus Christ. The point is I don't think the hardware can read such precise data from different ethnic groups when their features differ so wildly.

What? Why? The camera maps the face. Someone's ethnicity is irrelevant. It's not preprogrammed to read "Japanese" faces.

 

[teiresias]

Umm, you sit in front of the camera and the program does a calibration by mapping 500 reference points to your face (probably uses edge detection to find the edges of your face, and then to locat the eyes and the mouth, and possibly the nose) then it maps 500 reference points based on those boundaries and knowledge of general facial structure. They're not saying they've taken a specific person and painted 500 points on their face so the computer can map it - it's all done at a calibration screen during run-time.

 

[Varian]

Umm, you sit in front of the camera and the program does a calibration by mapping 500 reference points to your face (probably uses edge detection to find the edges of your face, and then to locat the eyes and the mouth, and possibly the nose) then it maps 500 reference points based on those boundaries and knowledge of general facial structure. They're not saying they've taken a specific person and painted 500 points on their face so the computer can map it - it's all done at a calibration screen during run-time.

Who exactly are you talking to?