Movements that went into Coraline's face was over 200,000

[DaveC426913]

I heard on the news a tidbit about the film Coraline. They said the number of combinations of movements that went into Coraline's face was over 200,000.

They made that sound like it was a huge number and a lot of work. But of course, I know they're talking about permutations, which dramatically decreases the necessary number of inputs.

I was trying to figure out (while driving) how few points of motion it would require to exceed 200,000 permutations, but I got stuck.

Say each point of movement (an eyebrow, tip of the nose, corner of the mouth) has only 5 positions (way down, down, neutral, up, way up). How many points must there be?

It's not 200,000 = 5^n is it?

That makes a mere 8 points of movement required to make 200,000 expressions.

 

[owlpride]

Say each point of movement (an eyebrow, tip of the nose, corner of the mouth) has only 5 positions (way down, down, neutral, up, way up). How many points must there be?

It's not 200,000 = 5^n is it?

That makes a mere 8 points of movement required to make 200,000 expressions.

This would be the correct setup for your assumptions. But I don't think it's quite as easy as that because not every possible combination of positions makes a meaningful facial expression.

 

[DaveC426913]

But I don't think it's quite as easy as that because not every possible combination of positions makes a meaningful facial expression.

It doesn't have to be meaningful; they're simply trying to claim a big number for the radio.

 

[owlpride]

Didn't they build physical models for the faces? Because in that case I doubt they would have built faces that don't correspond to meaningful expressions.

 

[DaveC426913]

Didn't they build physical models for the faces? Because in that case I doubt they would have built faces that don't correspond to meaningful expressions.

Perhaps, but I'll bet my paycheque they didn't build 200,000 physical models.

 

[owlpride]

Assuming they built the upper and lower halves of her face separately, they would get by with 1,000 physical models for 200,000 combinations. Or maybe they only built the jaws and animated the rest?

Replacement Faces

 

[DaveC426913]

Assuming they built the upper and lower halves of her face separately, they would get by with 1,000 physical models for 200,000 combinations. Or maybe they only built the jaws and animated the rest?

Replacement Faces

Why is it a good assumption that they would bulld the upper and lower halves of the face separately? If they're going to the trouble of making 200,000 possible facial expressions, I doubt they'd take such a huge shortcut as to pretend (erroneously) that movements of the jaw and mouth have no effect on the contours of the upper face. Those subtleties are kind of the whole point of going to that much trouble.

 

[DaveC426913]

So, what is the answer? With my proposal of a (very moderate) upper limit of 5 positions per point, how many points would the face have to have 200,000 unique combinations?

Was I right that the answer is 8?

I guess it has to be. Not sure why it seemed so difficult before. I guess because I started with the premise that it was going to involve a factorial rather than a simple exponent.

 

[owlpride]

Why is it a good assumption that they would bulld the upper and lower halves of the face separately? If they're going to the trouble of making 200,000 possible facial expressions, I dount they'd take such a huge shortcut as to pretend (erroneously) that movements of the jaw and mouth have no effect on the contours of the upper face. Those subtleties are knod of the whole point of going to that much trouble.

Did you look at the link I posted? It seems like a legitimate assumption because all pictures I have seen of Coraline faces only show half faces.

 

[DaveC426913]

Did you look at the link I posted? It seems like a legitimate assumption because all pictures I have seen of Coraline faces only show half faces.

I see your point.

 

[owlpride]

I heard on the news a tidbit about the film Coraline. They said the number of combinations of movements that went into Coraline's face was over 200,000.

I just realized that we should be more careful about positions vs movements. If we have x positions, there are x*(x-1) possible movements between two (ordered) positions. We would need about 450 different facial positions to get a total of 200,000 movements.

 

[maze]

One of the commonly used extensive models of facial animation has 68 parameters:
http://www.dsp.dist.unige.it/~pok/RESEARCH/MPEG/fapspec.htm

The parameters are real values between some maximum and minimum values, so the parameter space is a 68 dimensional box.