MakeHuman-Helpers

This page is used to explain the python CLI scripts only. The documentation of the blender plugins you will find here:

Please note that the plugins for 2.79 are also available for 2.80 and it is still under development. Some of the CLI scripts are included in the blender plugins as well.

Helpers are used for manipulating objects in MakeHuman Environment

All CLI programs use a common json-file for configuration, called:

data/mh_helper.json

It contains filenames of basemesh, mirror table and standard weights.

There are options to use different filenames in the program. All output is written to standard output so that no existing file is overwritten.

The configuration also contains the start vertices of the helpers.

I don't like JSON files with millions of lines for the weights. So the output of vertex/weights combinations is shortened to a 4 column table with 4 digits precision. You can change this by using -p and -c options.

Create mirror table

(Also included in blender plugin)

A small skript to create a mirror table of vertices from a symmetric (y-axis) wavefront (.obj) file. This table is used for symmetrizing objects (x <=> -x). Normally a deviation (MAXDIST) of 0.05 is allowed. The script starts with exact mirror, then 0.0001, 0.0002, 0.0005, 0.001, 0.002, 0.005, 0.01, 0.02 and 0.05. All vertices which have a mirrored partner are taken out of consideration. The Euclidian distance is used. A vertex near the mirror axis, which seems to be its own mirror, is correct to x=0.

If you want to reduce MAXDIST set deviation to e.g. 0.005.

create_mirror_tab.py [-h] [-b base_object] [-m MAXDIST]

The table contains all vertex numbers, mirror vertex numbers and orientations (l = left, r = right, m = mid)

Example:

..
18707 18690 l
18708 18708 m
18709 18691 l
..
19154 19157 r
19155 19156 r
19156 19155 l
19157 19154 l

Extract weights from weights file

A script which extracts groups from a given .mhw weight file

extract_weights.py [-h] [-f default_weights] [-t transpose_file][-p PRECISION][-c COLUMNS] RANGE

This skript extracts weight groups from a weight file, it is also able to transpose names according to a table. The transpose file then should contain lines of containing source group and destination group.

Example:

Consider the transpose file is called

data/default_bento.transpose

and looks like

#
# transpose file for default to bento
#
clavicle.L      mCollarLeft
clavicle.R      mCollarRight
eye.L   mFaceEyeAltLeft
eye.R   mFaceEyeAltRight
finger1-1.L     mHandThumb1Left
finger1-1.R     mHandThumb1Right
finger1-2.L     mHandThumb2Left
finger1-2.R     mHandThumb2Right
...

now use

extract_weights.py -t data/default_bento.transpose skirt

you will get

"mTorso": [
        [18002, 0.4441], [18003, 0.6071], [18004, 0.8432], [18005, 0.6358],
        ...
],

"mPelvis": [
]

Creation of symmetric weight groups

(also included in blender plugin, works like GROUP "=all=" in blender)

A script which creates a symmetric group or symmetrizes an existing one according to the name.

symmetrize_weights.py [-h] [-f default_weights] [-m mirror_table] [-p PRECISION] [-c COLUMNS] GROUP [ORIENTATION]

If GROUP contains .l, .L, left, Left, LEFT or .r, .R, right, Right, RIGHT the mirrored and the original groups are created.

All other groups are made symmetrical depending on a second parameter l or r to tell the program which side is preferred.

The GROUP "=all=" can be used to create identical weights on left and right side. Without ORIENTATION given the left side is copied to the right. In case of GROUP "=all=" also the rest of the original JSON file is copyied.

For calculation the program uses the mirror_table specified with -m or the file created by create_mirror_tab.py which should be copied to destination specified in the json-configfile.

Creation of a 100% symmetric wavefront file

A script which creates a symmetric geometry for a wavefront obj-file.

symmetrize_geom.py [-h] [-m mirror_table] OBJFILE [ORIENT]

Without mentioning the mirror_table the table out of the json-configfile is used. The orientatin should be l or r. l is default. The file is printed to stdout.

Example:

symmetrize_geom.py -m mirror_duck.txt duck.obj l >duck2.obj

Merging weights

When a new file was created it should be merged with the default weights file.

merge_weights.py [-h] [-f default_weights] [-p PRECISION] [-c COLUMNS] [-r] WEIGHTS

The skript merges the weights of a second file to the default weights. It can be used to replace groups or to add new weights to existing groups.

Add: e.g. a helper to the base mesh.

Replace: e.g. a beautified symmetrical version for the hair helper

Merging with an empty WEIGHTS file results in a version of the default weights using e.g. different PRECISION or COLUMNS values

Checking the weights in a weights file

To check if the weights will result to a normalized value of 1, the script weights_sum can be used

weights_sum.py [-h] [-f default_weights] [-L LOWEST] [-H HIGHEST][-s MINWEIGHTNUM] [-m MAXWEIGHTNUM]

All vertex numbers having smaller weights than LOWEST or bigger ones than HIGHEST will be printed. If only part of the file should be checked, the check may be reduced to a specific range of vertex number by using -s and -m option.

Normalizing weights

To get normalized weights of "bad" weights the skript normalize.py is used:

normalize.py [-h] [-f default_weights] [-s MINWEIGHTNUM][-m MAXWEIGHTNUM] [-p PRECISION] [-c COLUMNS] [-D DIFFERENCE]

Instead of writing only the incorrect lines this command creates the normalized values. At least one weight should be set for each vertex before, vertices with a weight of zero (yet undefined when you use the command weights_sum.py) are not processed.

Best to be used with a higher DIFFERENCE e.g. 0.005

Workflow: normalize a mesh

Normalizing a mesh. Just do a copy of the start mesh in a form we can compare with diff after processing. In this case we work with default_weights, the standard weight file of MakeHuman.

    merge_weights.py '' >/tmp/original_weights

Now let's do the normalization, the output only contains new weights:

normalize.py -D 0.005 >/tmp/normalized

Next merge the weights, replace mode is not used, because all weights not touched by normalization should stay as they are

merge_weights.py /tmp/normalized >/tmp/merged

Just to see what is processed, check the result by comparing with diff:

    diff /tmp/original_weights /tmp/merged

Workflow: create a new weight-file for a proxy

This workflow only works, if you are able to create a mirror table with blender or create_mirror_tab.py. Or you have made a mirror-table after applying the mirror for your new mesh. It must be able to create a table, where each vertex has a mirrored vertex. The base.obj in Makehuman fulfills this demand. For an own proxy best is to use a mirror operator in Blender when creating it.

Export a MakeHuman default character (male, female) with a proxy and the default skeleton supplied in mhx2 format, no subdivision used.
Import character into Blender, do not add or delete vertices (order is still original)
Now fix e.g. the area of the left kneecap by changing weights (e.g. by reducing the impact of upperleg on the lower part of the kneecap).
symmetrize the weights with the blender plugin (from left to right)
export with Blender exporter for .mhw format
change .proxy file so it refers to this weighting file, e.g. like this
```
max_pole 6  
vertexboneweights_file weightfile.mhw  
verts 0  
```
copy this file to the destinaton directory.

Append meshes to the basemesh

A command to append additional meshes to the base mesh. Designed for a base mesh in future editions.

append_base_mesh.py [-h] [-b default_base] [-p PRECISION] MESH

MESH (must be a wavefront .obj file) is concatenated to the base mesh (or default_base) to generate a new base mesh.

indigos33k3r / MakeHuman-Helpers