Wish List

[Manuel]

Manuel, if you are interested in the papers / research on this - I can give you pointers to more out there on body shape matching (again statistically based, but interesting & useful). Honestly though, the common andorgynous method has been proven to be (if not the best) a robust method of doing what MH does. I get annoyed whenever people ignore the abundant research and start spouting their ideas as gospel without backing it up with anything but their own anecdotal (and, in the case of pregnancy & biology, irrelevant) "proof".

--EK

I've read a lot of papers about this arrgument, but I'll take a look with pleasure.
Of course until yesterday I was sure that using androgyno was the best way, but thinking about chuck suggestion, I'm changing idea.
In particular the number of targets (take a look to my post at the end of page 2 in this thread) and the transition between
smooth to not-smooth are convincing arguments. What do you think about them?

[Eternl Knight]

I'll dig up some references for you on the body details.

On the subject of "smoothness" - chuck is correct about the ease which one can go from smooth->non-smooth in terms of both body shape & face. Please take a look at the average faces generated by FaceGen & the other research linked. These faces are "androgyn" and yet are "smooth". This is because the average (as in most common) face for males & females is smooth. I think the current "average head" being used is in MH is the difference between an idealized female face and the idealized male face. When you look around, most man are not square-jawed, strong-browed, and all angles. Most men are somewhere between the idealized man and idealized woman. Most women are closer to the idealized woman and are somewhere between the average man and idealized woman. In other words, I think that it is not the "concept" that is wrong (using an androgynous base), but the fact that your chosen base is not as close to the "real" average as is needed.

On the number of targets - I think there will be no real lessening of them. Good example is the way body mass is distributed when putting on muscle &/or fat. Men & women put bulk on in completely different areas, with differing amounts of muscle tone & or fat placement depending on the gender. In other words, for the most part you will need different targets for male/female anyway as there is not a "commonality" that can be used.

In other words, I believe (and have research to back it up) that the androgyn method is better than a base female with male morphs. The issue I feel for MH is not that it chose this method, but that it's current androgyn is more "male" than it needs to be because it is weighted toward "idealized male" rather than averaged between real male & real female.

--EK

[chuck_starchaser]

Obviously you've never lived in California ...

As a matter of fact I did, LA, for 3 years, around the time of the Iran hostage crisis.

It's funny that you mention that. I lived in southern Italy at the time working as an electronic spy for the USAF. I know things about the botched rescue mission that would make you cry.

I was working at a big company in LA, at the time, designing packaging and printed circuit boards for avionic electronics. I got the job of doing the tolerance analysis on some of the switches that were used in the panels in those helicopters, which supposedly were non-reliable and possibly to blame for that incident (one chopper's lights went on while the other chopper's pilot had the light amplifying goggles on, blinding him, or something along the lines, I heard). My analysis of tolerances showed both extremes of the tolerances to be non-functional: At one extreme, the switch would never close; at the opposite end, the switch would never open. But then I met an engineer at the gym, after work, who told me he had heard horrible things about that incident, but which he could not tell me about. So, I belive you, if in a WTF?? kind of way...

With all due respect, chuck, did you READ the research and/or use the demonstration application available in FacGen's site? I'm guessing not because your assertion that they have a different average female face and different average male face is incorrect.

Listen to the video you posted. 1 minute and 10 seconds into it.
http://www.youtube.com/watch?v=nice6NYb_WA

There is an "average face" for both male & female face. There are morphs (calculated using statistical methods based on the source data) that make the mesh more or less male/female. There is not a smooth continuum between male->female, but there are two smooth continuums male->average & average->female that are usable from this method.

Perhaps you're correct. What was said 1 minute and 10 seconds into the video was that average female and average male were used to produce male and female target sets, or however they phrased it. However, it boils down to the same thing as having a female-male continuum, mathematically speaking. The model of a face is based on A, but then you apply F-A to it, you get F. The same principle is not applicable to MH since we don't have an A-F target set. And in any case, to use A as the basis, and then use A-F and A-M target sets makes A superfluous; --might as well just have F and M.

I have studied the facial matching, facial generation, and even "body generation" research and commercial offerings and frankly all of them use an "average" mesh as their basis. This average mesh tends to have similar characteristics to that used by MH.

Appeal to Authority is a falacy; plain and simple. Such arguments as "this is the right way because it's the way everybody does it" have no weight whatsoever. Even such arguments as "this is correct because A says so, and A is an authority" can't hold any water, because if someone asks "and how do we know that A is an authority?", and the first persons says, "because B says that A is an authority, and B is an authority", then we can go on ad infinitum. Read Plato.

Regardless of all the biology you bring up (which you should know is completely meaningless in a 3D context),

I never knew that biology was 2D only, or did you mean something else? Just joking... Biology is as meaningful to MH as it is to anatomy. And it is very meaningful to anatomy. The male gender is produced by modification of the female gender; not the other way around. If you graph the level of testosterone in the placenta over the 9 months of a male baby's fetal growth, different curves correspond to different kinds of male physionomies and behaviors. (1) Those features that were developing at times of high testosterone levels develop more male; those that developed at times when there was a lower level of testosterone, develop less male. This variance is pretty much absent for female babies, unless they have a twin brother. I had a friend who was female, but really, REALLY gay; drooling at chicks that passed by sometimes. She was tall, strong, and her face looked a bit on the boy side of pretty boy. She had a male twin. This is a very common case in male/female twins: one or the other tends to be born gay, as well as having wrong gender physionomical features. High levels of testosterone in the placenta make a baby grow male, regardless of its chromosomes. This is something readily demonstrable with lab animal experiments. You can create male or female gay rats by supressing or injecting testosterone at the right time (when the opposite sex recognition brain structures are develping). And you can cause an otherwise female rat to have various distinctly male physionomical features by injecting testosterone in the placenta at various times during fetal growth. This has all been well researched. And the "natural" level of testosterone, when a male fetus is growing, is NOT constant; it varies; resulting in more than just a single "extent" of maleness, but many different *types* of malenesses. Thus, you can have different "directions" of male differentiation; different "dimensions" of maleness. And so it makes absolutely perfect sense for an anatomy modelling 3D application to model the way nature works, and represent "maleness" as an assortment of feature differentiations from femaleness.

NOTES:
(1) For those not aware of the fact, the growth of a fetus is a very tight and rigid schedule. Tissues and features develop at precise times from gestation. Lack of a particular nutrient in the mother's diet at a particular time during pregnancy often result in a tissue or organ being malformed. There's a condition, can't remember what it's called, but in which the baby is born kind of retarded, that results if the mother gets drunk on a particular date, like say 19 days into pregnancy, don't remember the exact number, due to alcohol preventing the formation of a particular section of the brain. If she gets drunk the night before, or the one after, no problem. It's not like a fetus grows slowly but all at the same time; no such thing; it's a rigid schedule.

[DAD]

{/quote] But then I met an engineer at the gym, after work, who told me he had heard horrible things about that incident, but which he could not tell me about. So, I belive you, if in a WTF?? kind of way...
[/quote]

It's long enough now that I can tell you what happened. I was on the line the night they went in. I was working analysis of my sub-element (which thankfully wasn't weather) and there was a hell of a big todo about getting weather intel for that area. The analyst and intercept operator covering weather that night forged a weather copy to get the war room off their backs. Their weather report said clear and calm ... the helicopters flew right into a sand storm! Needless to say those two guys vanished from the base real fast and were never heard from again! It had nothing to do with equipment failure ... but everything to do with people failure.

Sorry for the OT but sometimes it's fun to set the record straight.

[chuck_starchaser]

OMG ... Such a degree of criminal irresponsibility is hard to even fathom.

[Eternl Knight]

Listen to the video you posted. 1 minute and 10 seconds into it.

Yes, chuck - listen carefully to what is being said. It talks about creating the average faces for male & female individually such that the differences between them and the average face can be used to make a model more or less female/male. You completely ignore (purposely or not) the next sentence which is "other facial attributes, such as the fullness of the face, can be manipulated in a similar way". This is simply manual parameterization of the global morph-space, which is usually done by having people rate individual morphs with attribute values between 0.0 & 1.0 (i.e. this particular face is 0.4 bony, 0.9 african-american, 1.0 male, & 0.5 roman-nosed).

Unless you are suggesting that we create foundation meshes for fat, bony, african, middle-eastern, etc, etc in addition to the foundation male & female faces you propose - the argument falls flat. Please, actually read the research material and check your quotes in context before trying to argue about this. This is not something I found on the web and spent a night reading up on; I have done serious work in this area.

The same principle is not applicable to MH since we don't have an A-F target set. And in any case, to use A as the basis, and then use A-F and A-M target sets makes A superfluous; --might as well just have F and M.
A = Androngynous, F = Female, & M = Male

Actually, what we have is alot of morph targets taken from the "current" andro mesh which can be applied to either female & male meshes. As soon as we have a female & male mesh - we need to double them.

Appeal to Authority is a falacy

Quite true. However, in the scientific method, until such time as a theory is proven false - it can be used as the basis for other theories. Theories that are the most useful tend to be used to implement "practical technologies" (such as electricity generators, weather prediction, and anthropological face/figure generators). So far, I have provided links to a commercial implementation & the theory behind the current state of the art. You, however, have only provided your own anecdotal evidence without any backing it up with practical implementations or even ongoing research (let alone established & implemented technology).

Smart alec debate techniques are not a substitute for knowledge. You might be able to convince others that that your opponent is wrong, but that doesn't make YOU right.

Biology is as meaningful to MH as it is to anatomy.<snip/>

Yes, biology is meaningful; however the elements of biology you brought up (note, that was the important part of what I said before) IS meaningless. It doesn't matter how a body develops in the womb, especially when the project in question (i.e. MakeHuman) doesn't really generate really young figures well (nor can it using the current tech it relies on). Talking about how the body develops biologically from female to male in the womb means nothing in terms of vertices, bone matrices, and rigging. For example, it is somewhat silly to suggest that we should build a rig including breasts when the male rig is very different given the pectoral region will be flatter & made of muscle (& hence moving more with the underlying bones than the female breast).

It only makes sense for the anatomy modelling application to take serious consideration of biology on the level you discuss if the knowledge is used in such a way as to make the biology relevant. Given there is not even a "muscle simulator" underneath the mesh - I would posit that biology is only useful where it comes to the outside "appearance" of the mesh. The development & biological origins of the appearance are less important than the efficient generation & posing of the resulting figure.

In Summary...

To summarise, I think MakeHuman is on the right track for the development of human figure given the current state of the art & research in the matter. If I am mistaken, and MakeHuman is more a research project than an implementation project - than I'll stand back. If, however, MakeHuman's purpose is to take the current knowledge in this field and apply it to an open-source implementation of the ideas - I suggest we follow the current research as I have indicated.

NOTE: Body Mesh Research Link: here & here

--EK

[chuck_starchaser]

Listen to the video you posted. 1 minute and 10 seconds into it.

Yes, chuck - listen carefully to what is being said.

I did.

It talks about creating the average faces for male & female individually such that the differences between them and the average face can be used to make a model more or less female/male.

Yes. And in the next paragraph I said that if you average all faces, male and female and call it A; then average all female faces and call it F; then, when you move your slider to female, from the average face, A, you get F. Whatever other attributes you modify, from then on, are modifiers of F, not of A. Same difference.

You completely ignore (purposely or not) the next sentence which is "other facial attributes, such as the fullness of the face, can be manipulated in a similar way".

I ignored it because it's not related to the issue.

Unless you are suggesting that we create foundation meshes for fat, bony, african, middle-eastern, etc, etc in addition to the foundation male & female faces you propose - the argument falls flat.

YOU are the one who hasn't been paying attention to what I say. I did NOT propose to have male and female meshes. I've been proposing ONE mesh: female.

Please, actually read the research material and check your quotes in context before trying to argue about this.

I read them, already; you mean that I should read them again?

This is not something I found on the web and spent a night reading up on; I have done serious work in this area.

I never implied anything about your background.

The same principle is not applicable to MH since we don't have an A-F target set. And in any case, to use A as the basis, and then use A-F and A-M target sets makes A superfluous; --might as well just have F and M.
A = Androngynous, F = Female, & M = Male

Actually, what we have is alot of morph targets taken from the "current" andro mesh which can be applied to either female & male meshes. As soon as we have a female & male mesh - we need to double them.

I wasn't suggesting that we have M and F meshes. What I said is that what they are doing is mathematically equivalent to having M and F meshes. I wasn't endorsing it; simply pointing out that your putting so much emphasis on them having an A (average face) is misplaced emphasis, since the way the M-F slider is implemented it creates the equivalent of having M and F meshes.

Appeal to Authority is a falacy

Quite true. However, in the scientific method, until such time as a theory is proven false - it can be used as the basis for other theories. Theories that are the most useful tend to be used to implement "practical technologies" (such as electricity generators, weather prediction, and anthropological face/figure generators). So far, I have provided links to a commercial implementation & the theory behind the current state of the art. You, however, have only provided your own anecdotal evidence without any backing it up with practical implementations or even ongoing research (let alone established & implemented technology).

Totally agreed; but it doesn't apply here, since none of the research (that you showed me links to, anyways) ever presents the use of an average face being the "best way" as the subject of research. They merely say "we did this". The implication, to me, is that it seemed like a good idea; so it was done that way. Science is about asking questions, formulating hypothesis, and trying to prove them.

Smart alec debate techniques

Such as ... ?

are not a substitute for knowledge.

Knowledge is very powerful. If you're saying you know the truth and I'm wrong but that my debate techniques are so damn good that I can uphold falsehood and keep your truth at bay, well, thank you!

You might be able to convince others that that your opponent is wrong, but that doesn't make YOU right.

That might be true, but then, what are you going to do about it? There's only one way to establish truth and that is by time-honored dialectics.

Biology is as meaningful to MH as it is to anatomy.<snip/>

Yes, biology is meaningful; however the elements of biology you brought up (note, that was the important part of what I said before) IS meaningless. It doesn't matter how a body develops in the womb, especially when the project in question (i.e. MakeHuman) doesn't really generate really young figures well (nor can it using the current tech it relies on). Talking about how the body develops biologically from female to male in the womb means nothing in terms of vertices, bone matrices, and rigging.

It means everything. I did not suggest we model a fetus and simulate its growth. But the adult anatomy is quintessentially decided during fetal development. You can exercise or be sedentary; you can eat a lot of pasta, or you can eat steak, but all you'll do is achieve changes that are temporary and conditional, and reversible; but your quintessential anatomy is preserved, and that was decided during fetal growth. In fact, come to think of it, targets could be classified into temporary and permanent. Musclarity, bodyfat, postures and expressions would be temporary. Anyways, if you could show that there are many factors beyond fetal development to consider, I might consider your argument; but in fact the biology in the womb is the sole determinant of our quintessential individuality. The fact that male fetuses go through a roller-coaster of testosterone exposure during fetal development and females don't is very significant, in the sense that if we wanted to have an F-M slider we'd actually need a large number of sliders: F-M1, F-M2, F-M3; --where M1 is the average male among those who got higher testosterone in day 72, M2 are those males who got more testosterone in the womb on day 73, etceteras.

For example, it is somewhat silly to suggest that we should build a rig including breasts when the male rig is very different given the pectoral region will be flatter & made of muscle (& hence moving more with the underlying bones than the female breast).

Again, I've been talking about head and face all along. I would never suggest we have targets to turn female genitalia into male genitalia.
(And before someone tells me that I said I would NOT do the same with the body, read my lips: I advocated nothing about the body; period. Haven't thought about it.)
And if I thought about it, the F-M slider MH works well enough for me. My big concern is faces, really. The goal to strive for is that with a minimal number of targets you can produce ANY face, male or female, exactly. If you used MH to model Christina Aguilera and Brittney Spears, I should be able to tell which is which.

It only makes sense for the anatomy modelling application to take serious consideration of biology on the level you discuss if the knowledge is used in such a way as to make the biology relevant. Given there is not even a "muscle simulator" underneath the mesh - I would posit that biology is only useful where it comes to the outside "appearance" of the mesh. The development & biological origins of the appearance are less important than the efficient generation & posing of the resulting figure.

Posing is as far away from the subject matter as anything gets. The issue here is that males and females are pretty much identical in feature set, but different in the expressions of that feature set; and while females show many hereditary variations, males show similar hereditary variations PLUS a huge combinatorial set of maleness variations due to the time curve of testosterone levels in the womb during fetal development. So there's more feature variation in males than in females. (Something that females seem terribly self-conscious about. Freud was wrong: it wasn't penis envy; it was variability envy.)

In Summary...To summarise, I think MakeHuman is on the right track for the development of human figure

I'll say it one more time, just in case... I'm talking about faces.

given the current state of the art & research in the matter.

I'll say this again too: Show me research that addresses the specific issue of why androgyn should be the starting point.

If I am mistaken, and MakeHuman is more a research project than an implementation project - than I'll stand back. If, however, MakeHuman's purpose is to take the current knowledge in this field and apply it to an open-source implementation of the ideas - I suggest we follow the current research as I have indicated.

That's like saying that the developers of MH shouldn't think for themselves, but blindly follow whatever the authorities in the subject have decided is right.

NOTE: Body Mesh Research Link: here & here

Neither of these links discusses why androgyn should be the starting point.

[Eternl Knight]

Show me research that addresses the specific issue of why androgyn should be the starting point

I have. You have not read the documents and are quoting one video out of context. Until you can back up your claims with research, I have more facts on my side. All your semantic arguments to the side, you have proven nothing and made a variety of assertions you cannot back up with anything but your own opinion. At almost every step of the way, you have ignored or twisted what I have said & yet provided

I have provided research material and a proven commercial application that use androgynous base meshes (for facial characteristics, which you claim are all you are talking about). Until you back up what you say with a successful implementation and/or research saying otherwise, you have nothing to say I am interested in.

Manuel: I am finished with this topic. If you are interested in further research & implementation links - let me know by PM. Also, please let me know if this is a research project or an implementation project as I will refrain from debating the merits of 'proven techniques' if you are interested more in the research side.

--EK

[chuck_starchaser]

Show me research that addresses the specific issue of why androgyn should be the starting point

I have. You have not read the documents and are quoting one video out of context. Until you can back up your claims with research, I have more facts on my side. All your semantic arguments to the side, you have proven nothing and made a variety of assertions you cannot back up with anything but your own opinion. At almost every step of the way, you have ignored or twisted what I have said & yet provided

I have provided research material and a proven commercial application that use androgynous base meshes (for facial characteristics, which you claim are all you are talking about). Until you back up what you say with a successful implementation and/or research saying otherwise, you have nothing to say I am interested in.

Manuel: I am finished with this topic. If you are interested in further research & implementation links - let me know by PM. Also, please let me know if this is a research project or an implementation project as I will refrain from debating the merits of 'proven techniques' if you are interested more in the research side.

--EK

Man, calm down; this is not a funeral...
Assuming the worst: that I'm completely wrong, yet through devious argumentative techniques have somehow managed to mislead Manuel down a terribly wrong path, the worst consequence of such a scenario would be a discovery of my error after some wasted time, and a learning experience for all involved. And I'm sure the experimental new code and data will be implemented onto an svn branch of the main repository, and only merged if and when it proves its worth; so it's not like MH is about to be destroyed or something.

Show me research that addresses the specific issue of why androgyn should be the starting point

I have.
...............................................
I have provided research material and a proven commercial application that use androgynous base meshes ...

I said several times that it's NOT enough to show that others are USING androgynous base meshes, but that to prove me wrong you need to show me reasearch into the specific subject of why using an androgynous average face is better than using a female average face as base mesh (or any other alternatives that were looked at, for that matter). Your saying "I have", I believe it to be a lie; --nothing more; nothing less. And the second sentence I quote above proves it, IMO.

Now, I do realize I'm putting a heavy burden of proof on you; but your "argument techniques" left me no choice: You just kept going back to appealing to authority, again and again and again; which would be a falacy all on its own, to begin with; but it's made even more so by the fact that none of those commercial apps or research projects specifically addresses the question of whether androgynous average, or something else, is the best choice of base mesh, --much less why; they all seem to have simply assumed that choice to be "common sense", and to have never questioned it. So, however great their authority and success may be, it is offered as proof, here, ***out of context*** ((be this an example of a rightful and correct use of the term, by the way)).

For what I've seen so far, I'm probably the first person in the whole Milky Way galaxy to ever come up with the observation that "androgyn" is a red herring, and that probably the average female should be used as the base mesh, instead. Chances are none of those researchers and commercial product developers ever gave this question any serious thought; but just followed the tradition set by everyone that came before them. I bet you any money it is so. The concept of "androgynous average" in face analysis is probably older even than computers. A perfect candidate for the type of idea that becomes so ingrained that nobody ever questions it, until one day some idiot like me cries out that the emperor has no clothes on.

The very fact that you argue for androgynous average as base mesh with so much emotional investment and so much recourse to pointing out what others do, is circumstantial but strong evidence for what I'm hypothesizing in the last paragraph. It just shows how even a smart and well informed individual such as yourself can become prey to the common assumption that "if most people do X, then X must be right". Plus an irrational, emotional reaction to any ideas threatening that assumption.

Such resistance to questioning of the status quo is precisely what typically prevents many new ideas from being considered, --or even conceived. It shows a typical example of what can happen, and why probably nobody else ever wondered (or dared to wonder aloud) whether androgynous average is indeed the best choice for base mesh, or not, --and why--, in the first place. Most people abhor argument so much they would rather supress their own questioning minds than risk facing the kind of trouble I just got myself into.

you have proven nothing and made a variety of assertions you cannot back up with anything but your own opinion

This is true, but to throw this on my face you need to at least tell me what kind of evidence you would be willing to consider. The way I see it, you sound like when physicists criticized the Theory of Special Relativity citing the lack of empirical proof. Empirical proof came, in time; but Special Relativity was more of a turning upside-down of the existing assumption of time being some kind of universal, synchronous clock. It questioned whether the concept of "simultaneous" was actually objectively definable, or subjective to any given, moving frame of reference. Such questioning did not call for much in the way of empirical proof. His theory helped explain certain curious observations, like the bending of light near the sun; or like the fact that the speed of light through a moving medium, such as water flowing through a pipe, did not seem to be affected by the speed or direction of the fluid's motion. But there were other, competing hypothesis, at the time. Special Relativity was a revolutionary way of thinking, mostly; though it implied some consequences about clocks aboard fast planes circling the equator that could only much later be put to the test. Quantum Indeterminacy is another example of a theory that came out without much in the way of a log of experiments attached.

Naturally, you're going to accuse me of comparing myself to Einstein and Heisenberg, now...

Anyways, your saying "... anything but your own opinion", as if implying that 'opinion' is a worthless substance, is inconsistent with the argument you made for science and technology and whatnot. By means of the Socratic method of Dialectics (by whatever name they call it), people try to arrive to the truth of a matter of disagreement. This is what underlies the practice of the Scientific Method and what goes on at Engineering meetings everywhere, which is what permits the evolution of Technology. Opinion is only "nothing but opinion" precisely when people refuse to discuss and debate; --or quit from a discussion having neither prevailed with their view, nor given the point, nor somehow agreed to postpone resolution until an experiment can be made that both parties agree should settle the issue. So, by saying "I am finished with this topic" you are in fact making ***your*** opinion into "nothing but... opinion".

[Manuel]

First thing, guys, please calm, less OT and more objective.
I appreciate this discussion (but not the flame), because we have seen a series of important points:

• The actual mesh can be improved.
  EK has observed that " The issue I feel for MH is not that it chose this method, but that it's current androgyn is more "male" than it needs to be because it is weighted toward "idealized male" rather than averaged between real male & real female.", while Chuck prefer a full female base mesh instead actual mesh.
  
• Modify a "smoth" surface to transform it in a "no smooth" surface is moooore simple than the contrary.
  In particular, smooth -> Nsmooth can be obtained easily with actual targets technology, while
  Nsmooth -> smooth require a specific (and not so simple) algorithm
  
• We all agree that base mesh should be more smooth. So the difference between EK and chuck point of view is not so big.
  In my opinion it's only matter of measure: we have just "more feminine mesh" vs "full female mesh".

There are thing to be studied:

• The number of targets can be reduced?
• Androgyne is really needed?
  Assumend A = androgyn, F = female, M = male
  we can have two cases, that I've quickly sketched below. In case (a) androgyne is retundant, while in (b) it's needed.

I like the experimental style. So if we found results in contrast with previous works, and our results are better....: cool! We publish our own paper !!!