Parametric modeling?

[AtheneNoctua1]

Have returned to Makehuman after eight months and find that some wonderful changes have been implemented!

Here's my suggestion:

When the user moves the slider to adust the parameters, the program stores the number value of the morph/warp in the .mhm document. If we could see and adjust that number value in the GUI, we could make changes more precisely. Also if we wanted, we could make sure the numbers on the right side and left side, such as for hands and ears, are exactly the same. The reason this is important: The X-mirror in Blender doesn't work unless the two sides across the model's X-axis are exactly the same.

Thanks again for all your hard work.

[duststorm]

MakeHuman has symmetry buttons in the top toolbar for precisely this.
Numeric input has been discussed oft and long, and the conclusion is that manually entering values in the [0, 1] range just does not make much sense. People asking for numeric input usually want to specify empricial sizes, like weight in kg or length in m. This is currently not possible with MH.

[zoiner]

People asking for numeric input usually want to specify empricial sizes, like weight in kg or length in m. This is currently not possible with MH.

Just what I would like. I want to generate models of real people of known dimensions to place them in scale simulations of working environments and task equipment. It can be a real pain trying to nudge the slider to exactly the point I want.

Alternatively (ignorance here) can I save an approximation in a custom file and then edit the parameters in the file using a text editor?

[duststorm]

Alternatively (ignorance here) can I save an approximation in a custom file and then edit the parameters in the file using a text editor?

Sure, you can save your model as .mhm and tweak the values with a text editor, then load it back in MH.
Note that only slider values of sliders you have touched (not their default positions) will be included in the file.

[zoiner]

Great, thank you!

[zoiner]

Great, thank you!

Ah, just got V1.0.0. Not so great

The mhm file is stored in relative units (0-1) not absolute values, so it is necessary to have some form of look-up table for each dimension!

Are the Scales linear? So If i know the end points and the corresponding absolute values can I shift them?

Hold on the absolute value for each will vary dependent on others so the stature range will vary dependent on the gender value, for example.

Is there an option for saving a file in absolute units?

[duststorm]

The mhm file is stored in relative units (0-1) not absolute values, so it is necessary to have some form of look-up table for each dimension!

That's what I said from the start.

Are the Scales linear?

Yes, it's a linear morph.

Hold on the absolute value for each will vary dependent on others so the stature range will vary dependent on the gender value, for example.

Yes, it's a linear system, but a very large one.
Additionally the linear equation that relates real-world measurements to targets will be an over-specified/determined one (a heavy 10 year old can weigh as much as a thin 13 year old, for example).

Is there an option for saving a file in absolute units?

Currently not, no. MakeHuman does not work with real world units. Adding support for such a thing is not an easy task.

[duststorm]

If this interests you, here is an excerpt from a chat about it:

[19:58:13] <Duststorm> measures are modifiers as well
[19:58:28] <Duststorm> take a look at the 0_modeling_...measurement plugin
[20:00:27] <Duststorm> see displayToData() function
[20:00:37] <Duststorm> there is no way to directly apply a numerical value
[20:00:44] <Duststorm> you need to do it by trial and error
[20:01:02] <Duststorm> apply modifier at certain amount, measure the mesh, try again
[20:02:12] <piccir> yes I was studying that and the way it works is to modify the vertex of the body part until that measure is reached, I was just looking for a high level function, as you load a mhm file
[20:03:05] <Duststorm> nothing like that is implemented
[20:03:17] <Duststorm> because we have an artistic approach
[20:03:24] <Duststorm> not a mathematical model
[20:04:51] <Duststorm> you could probably speed it up by building a search matrix of metadata (precalculated measurements) over the target space (all target combinations)
[20:05:37] <Duststorm> but you would probably soon notice that it isnt a 1-1 mapping
[20:05:47] <Duststorm> and that for many input domains there are multiple outputs
[20:06:14] <Duststorm> eg torso circumference of a thinner male == that of a heavier woman
[20:06:26] <Duststorm> and there would be many such clashes
[20:10:01] <piccir> I've already noticed that if you set the macro, the measures are pretty realistic, is just that sometimes I want to cover more subject the i need to put some measures, the measure plugin I think work well is just that the setting the micro measures is different than the macro.....
[20:11:11] <piccir> clearly if I use the -1 to 1 values as absolute I'm screwed
[20:13:05] <Duststorm> I have never really formally looked at the ranges covered
[20:13:32] <Duststorm> I know that the measure targets are simple targets
[20:13:37] <Duststorm> they dont depend on macro targets
[20:13:57] <Duststorm> so they always add or deduct the same measure units from your model
[20:14:09] <Duststorm> irrelevant what the macro is set to
[20:14:15] <Duststorm> so that simplifies things a lot
[20:14:58] <Duststorm> so if you record the measures of the macro combinations
[20:15:15] <Duststorm> and calculate the offsets of measurment targets
[20:15:31] <Duststorm> then you know what domain you can model in
[20:16:06] <piccir> That is nice! (I was just asking that)
[20:18:29] <piccir> Have you thought about the opposite process?? From a overall body decompose it in many targets?
[20:25:41] <piccir> At this point can be useful a plugin that given real measures in cm can build the subject with that measures
[20:29:27] <Duststorm> its not enough
[20:29:33] <Duststorm> as I said before
[20:29:46] <Duststorm> but you would probably soon notice that it isnt a 1-1 mapping
[20:29:52] <Duststorm> and that for many input domains there are multiple outputs
[20:29:59] <Duststorm> eg torso circumference of a thinner male == that of a heavier woman
[20:30:33] <Duststorm> you could create your own set of targets
[20:30:38] <Duststorm> eg using maketarget
[20:31:03] <Duststorm> per body part, directly on the basemesh that 1-1 map to your measurements
[20:31:19] <Duststorm> but you will probably be creating very bland and generderless characters
[20:35:01] <Duststorm> only building a character with as input measures will probably result in a character as bland as: http://3.bp.blogspot.com/_pFGwpjnZD-g/S ... 100038.JPG
[20:35:11] <Duststorm> (but full body of course)
[20:35:37] <Duststorm> and then still, you'd have to use very fine measurements as input
[20:36:00] <Duststorm> (to be able to tell if the user wants a male or female, for example)
[20:36:27] <Duststorm> breast circumference wouldnt suffice of course
[20:37:11] <piccir> Yes that could happen, but it depend the decomposition that you're using for the parameters
[20:38:00] <piccir> My question is more like: how to use all the existing targets with a different parametrization
[20:39:38] <Duststorm> build your own index
[20:39:42] <Duststorm> as I already told you
[20:39:56] <Duststorm> apply targets, measure, store, repeat
[20:40:19] <piccir> yep that the way to go
[20:40:19] <Duststorm> then you can input your desired measures in your search matrix (index)
[20:40:31] <Duststorm> and it will return the matching target combinations
[20:40:57] <Duststorm> mh targets are linearly applied
[20:41:10] <Duststorm> so you only need to calculate the extremes (applied with weight 1)
[20:41:19] <Duststorm> and 0
[20:41:22] <Duststorm> or -1
[20:41:58] <Duststorm> and then its solving a very large linear system equation
[20:42:22] <Duststorm> but I warn you, its not going to be deterministic
[20:42:58] <Duststorm> one combination of measurements can have lots of different target combinations
[20:43:31] <Duststorm> you could analyse your search matrix, and either 1) determine what extra inputs you need to make the searches deterministic
[20:44:14] <Duststorm> 2) drop all targets that cause conflicts so you are using a more limited set
[20:44:22] <piccir> if you put the problem in that way is hard to have one solution: it's what is called an ill posed problem...
[20:44:27] <Duststorm> 2 will again result in the dress doll effect
[20:44:35] <Duststorm> less variability
[20:46:29] <piccir> I think it can be treated as a machine learning problem too......
[20:47:49] <Duststorm> yes
[20:48:07] <Duststorm> but as I said, you will have to deal with the conflicts
[20:50:29] <piccir> yes it's not an easy job
[20:51:45] <piccir> Thank you for the explanation, very helpful

... (some other related chat)

[17:27:39] <Duststorm> the only way to do this in a manageable way is probably introduce some n x n matrix with weight, height, ... metrics precalculated
[17:27:49] <Duststorm> in which it can perform lookups
[17:28:06] <Duststorm> if you store all extremes of all macro targets that matter
[17:28:33] <Duststorm> its pretty fast to calculate the values using linear interpolation
[17:28:40] <Duststorm> the question is, is this worth the effort?
[17:28:49] <Duststorm> this is a considerable task
[17:29:10] <Duststorm> it might be worth it if this system will be useful for the measure plugin too
[17:29:17] <Duststorm> to work the other way around
[17:29:24] <Duststorm> from metrics to targets
[17:29:56] <Duststorm> but as I tried to explain marco, this is a linear system equation that is over-specified
[17:30:01] <Duststorm> with many possible solutions
[17:30:19] <Duststorm> metrics -> target combinations, I mean
[17:30:30] <Duststorm> target combinations -> metric is deterministic

If someone is interested in contributing and working on this, the help is welcome!

[SamA]

You can use the tools in the Measure tab to make adjustments based on real world units.
I emphisize make adjustments because they are not for modelling, you should rough out the proportions in the other modelling tabs first, then use Measure to fine tune to required values.

[Artist]

MakeHuman has symmetry buttons in the top toolbar for precisely this.
Numeric input has been discussed oft and long, and the conclusion is that manually entering values in the [0, 1] range just does not make much sense. People asking for numeric input usually want to specify empricial sizes, like weight in kg or length in m. This is currently not possible with MH.

Although not hardly everyone will need a numeric input for those relative values from -1 to 1, the option should be there for those that do. It should be a trivial thing to include an optional numeric entry control to adjust the same number that is adjusted by slider. So why not have it there for the few who need it? Every slider should have this.