How to use this face data

[fierce]

Im trying to extract models from a game that Noesis supports partially, Im a bit new regarding 3d model reverse engineering so I have no idea how the vertices/normals/faces data is organized on most formats.
On this files the vertices, normals and textures seems to be pretty straight forward, but I'm not having the same luck with the faces.
On a model I have this for example:

(After being exported from Noesis)
I know for a fact that those numbers represent the right vertices order. (I've checked manually)
Noesis saves those faces like this:
f 1/1/1 2/2/2 3/3/3
f 4/4/4 3/3/3 2/2/2
f 4/4/4 2/2/2 5/5/5
f 6/6/6 5/5/5 2/2/2
f 6/6/6 2/2/2 7/7/7
f 8/8/8 7/7/7 2/2/2
From here its easy to figure out how the faces are built,
but when I check those faces data on the file I have:
1 2 3 6 -1 6 2 5 4 -1 4 2 7 8 -1 (0000000100020005FFFF0005000100040003FFFF0003000100060007FFFF) (I've incremented the numbers by 1 so it makes more sense)
And I have no idea how to put that faces data together to export it to an .obj file (for example).

This seems to be a case of triangle strip but I have no idea how that data is organized, can you guys give me a hand?

Thank you in advance

[herbert3000]

Could you please post the coordinates of the vertices?
The faces data (1 2 3 6 -1 6 2 5 4 -1 4 2 7 8 -1) only makes sense to me if I reorder the vertices like this:

[shakotay2]

Im trying to extract models from a game that Noesis supports partially,

What's the name of the game? "Secret"?

but when I check those faces data on the file I have:
1 2 3 6 -1 6 2 5 4 -1 4 2 7 8 -1 (0000000100020005FFFF0005000100040003FFFF0003000100060007FFFF) (I've incremented the numbers by 1 so it makes more sense)
And I have no idea how to put that faces data together to export it to an .obj file (for example).

Hmm, interpreting FFFF as '1'?

For me it's
0000 0001 0002 0005 FFFF -> 1 2 3 6 (FFFF) ; ignoring the FFFFs
0005 0001 0004 0003 FFFF -> 6 2 5 4 (FFFF)
0003 0001 0006 0007 FFFF -> 4 2 7 8 (FFFF)

This seems to be a case of triangle strip

Yep, should be quads which are split up into faces by Noesis.
But splitting up into faces would mean for the quad 6 5 2 4 -> 6 5 2 and 2 4 6 or something like that.

Could you please post the coordinates of the vertices?

Strongly recommended; or better:
the whole Noesis output or at least all the faces and the corresponding data block from the file.

[fierce]

Im trying to extract models from a game that Noesis supports partially,

What's the name of the game? "Secret"?

but when I check those faces data on the file I have:
1 2 3 6 -1 6 2 5 4 -1 4 2 7 8 -1 (0000000100020005FFFF0005000100040003FFFF0003000100060007FFFF) (I've incremented the numbers by 1 so it makes more sense)
And I have no idea how to put that faces data together to export it to an .obj file (for example).

Hmm, interpreting FFFF as '1'?
For me it's
0000 0001 0002 0005 FFFF -> 1 2 3 6 (FFFF) ; ignoring the FFFFs
0005 0001 0004 0003 FFFF -> 6 2 5 4 (FFFF)
0003 0001 0006 0007 FFFF -> 4 2 7 8 (FFFF)

This seems to be a case of triangle strip

Yep, should be quads which are split up into faces by Noesis.
But splitting up into faces would mean for the quad 6 5 2 4 -> 6 5 2 and 2 4 6 or something like that.

Could you please post the coordinates of the vertices?

Strongly recommended; or better:
the whole Noesis output or at least all the faces and the corresponding data block from the file.

Naruto Shippuden: Ultimate Ninja Storm 1.

I was ignoring the FFFF aswell on the beginning, they sounded more like a "separator" to me, although it also might be treated as -1, but has I said, I'm new

I'll send you the file with one object/mesh, it's from a file available on the online demo, downloadable on the PlayStation website so I think its ok to send it. (I've slapped a NUCC header to it just so it can be read by Noesis).

Vertex data from the file:

1628.789 1.410156 -3.90326
1515.981 -40.51709 -3.903336
1623.279 -0.9135742 88.91228
1448.056 -85.29028 136.3096
1515.981 -40.51709 144.5438
1604.066 -7.868164 124.5788
1377.035 -131.3052 68.12234
1388.034 -124.4492 -3.903328

And you were right herbert3000, I was using the data from the exported file from Noesis that's why I got that vertex order, totally forgot that Noesis swapped their order. I checked the vertex order directly from the file and it matches the one you pasted. Thanks a lot.

It is helpful to know how to handle this data so I can write an extractor/converter or a Noesis script and share it with you.

Thanks a lot for the help guys.

On a side note I already found out what data Noesis isn't reading properly on some files. Its the number of bytes that correspond to each (vertex coord.)+(normal)+(?????)+(texture coord.), not sure what the third one is.
On offset 6E and 6F, there are two int8's that tells that each vertex coord. (6E) and normals+????+texture coord (6F) corresponds to X bytes.
Example on this file:
Offset 6E = 0x06 (6) = Vertex coordinates are represented in 3 floats (6 x 2 bytes?)
Offset 6F = 0x12 (18) = Normals + ???? + Texture coordinates.
This one doesnt make much sense but it works if I change it to 0x10 (16), that would mean 16 bytes:
6 bytes for normals (3 half floats) + 6 bytes for ???? + 4 bytes for texture coordinates (2 half floats)
I'll give you the example for the first vertex:

1: 44CB 993F 3FB4 8000 C079 CF01 B592 3B7F 19B4 3C00 FFFF FFFF 342D 342D
2: 44BD 7F64 C222 1180 C079 D041 B735 3B24 0000 3C00 FFFF FFFF 3800 3800
3: 44CA E8EE BF69 E000 42B1 D316 B588 3B81 0000 3C00 C8C8 C8FF 37C8 2B9E
4: 44B5 01CB C2AA 94A0 4308 4F3F B862 3AB1 170F 3C00 C8C8 C8FF 3BE0 356E

Translated:

1:
v 1628.789 1.410156 -3.90326
vn -0.3481445 0.9370117 0.002785
(3C00 FFFF FFFF) Dont know what this corresponds to
vt 0.2609863 0.2609863
2:
v 1515.981 -40.51709 -3.903336
vn -0.4504395 0.8925781 0.
(3C00 FFFF FFFF) Dont know what this corresponds to
vt 0.5 0.5
3:
v 1623.279 -0.9135742 88.91228
vn -0.3457031 0.9379883 0.
(3C00 C8C8 C8FF) Dont know what this corresponds to
vt 0.4863281 0.059509
4:
v 1448.056 -85.29028 136.3096
vn -0.5478516 0.8364258 0.001723
(3C00 C8C8 C8FF) Dont know what this corresponds to
vt 0.984375 0.3393555

Note: The third value on vertex textures is always 0.000000 after being exported by Noesis and I dont find any 0 values on the data so I'm assuming its not necessary in this case?

Can you guys help me figure out what those 6 bytes correspond to? Is it another normal coord? That would make it have 544 normals instead of 272, no?

Thank you in advance for any help

[shakotay2]

but has I said, I'm new

Just kidding? As far as I can see your knowledge is somewhat advanced.

Can you guys help me figure out what those 6 bytes correspond to? Is it another normal coord?

"Square sum" of components of normals vector should give 1.0
This isn't true for 1.000000 -131008.000000 -131008.000000

v 1628.789307 1.410156 -3.903260
vn -0.348145 0.937012 0.002785
#? 1.000000 -131008.000000 -131008.000000
vt 0.260986 0.260986
v 1515.981445 -40.517090 -3.903336
vn -0.450439 0.892578 0.000031
#? 1.000000 -131008.000000 -131008.000000
vt 0.500000 0.500000
v 1623.279053 -0.913574 88.912277
vn -0.345703 0.937988 0.000031
#? 1.000000 -9.562500 -9.992188
vt 0.486328 0.059509
v 1448.056030 -85.290283 136.309555
vn -0.547852 0.836426 0.001723
#vn 1.000000 -9.562500 -9.992188

There are only few different 6 bytes groups - first word is always 3C 00, last byte always FF.
So I think you can safely ignore them.
You have v, vn, vt and the faces. You don't need more, do you?

[fierce]

but has I said, I'm new

Just kidding? As far as I can see your knowledge is somewhat advanced.

I'm new to game dev and reverse engineering, been coding for some years though.

How do you handle the data to get (-131008.000000 -131008.000000) out of (FFFF FFFF)? It's not very important in this case I just want to know because Noesis is saving some vertex textures like that and I would like to understand how its handling the data.

There are only few different 6 bytes groups - first word is always 3C 00, last byte always FF.
So I think you can safely ignore them.
You have v, vn, vt and the faces. You don't need more, do you?

You're right, thank you.

Now I only got a few issues left. I figured out almost all of the data on the header file. There seem to be vertex groups there. It's probably of common usage and it makes perfect sense. I just haven't figured out exactly how Noesis handles that data. Here are the vertex groups explained:


PS: I've edited the values of the assignments. They used to be: 06 1C - 06 1C - 06 1C - 06 1C. I've changed them to 06 14 - 06 14 - 06 14 - 06 14 as you can see on the image.

The obvious questions are:
How does the data assignment works on e)?
The 06 seems to be 6 groups of 2 bytes (3 groups of 4 bytes) representing the vertex coordinates.
Why does the second assignment 1C (28) when there are only 20 bytes (in groups of 2 bytes)
Any idea what those 2 groups that I couldn't identify mean?

And example of some vertex data from each group just to help visualize it:




Thank you in advance for any help

[shakotay2]

How do you handle the data to get (-131008.000000 -131008.000000) out of (FFFF FFFF)?

I used this formula: f = ((h&0x8000)<<16) | (((h&0x7c00)+0x1C000)<<13) | ((h&0x03FF)<<13) ;
from the web for the half float (h) to float (f) -conversion.
The result is 0xc7ffe000 for h = 0xFFFF. Maybe this case is not handled by the formula.
edit: oh, yes, my bad!

Finally, if the exponent is 31 and the mantissa is not zero:
value = ±NaN (Non a Number)

(To check whether my selfwritten I3E750_to_float() function is ok I used http://www.h-schmidt.net/FloatConverter/IEEE754.html. For 0xc7ffe000 this also gives -131008.000000)

[fierce]

Thank you.

Any idea on those assignments? The 06... part
I'm checking different files at the moment to see if I can find any pattern. So far I only got:
When 06 12 (06 18) is set, there are usually 16 bytes for vn/vt data. (6+6+4)
When 06 1C (06 28) is set, there are usually 20 bytes for vn/vt data. (6+6+4+4 ?)
When 06 22 (06 34) is set, there are usually 20 bytes for vn/vt data. (6+6+4+4 ?)
In all those cases I think the 06 refers to the vertex coord data (3 floats of 4 bytes each).

I think that Noesis only messes up when there are more than 1 or 2 vertex groups.
If I change the second byte to the correct amount of bytes on the vn/vt data I get everything right except the vertex textures.
I'm also trying to write a Noesis plugin to try to extract the data properly but its my first one so it might take a while, and I have no idea how to get the data out properly if I kind find the association of the vn/vt data.

[shakotay2]

I'm also trying to write a Noesis plugin to try to extract the data properly

Does Noesis use a (somehow fitting) python plugin for Ninja Storm files? (If so it would be imho easier to change this plugin for your needs.)

To my knowledge the vertex group (numbers) are needed for skinning.

Seems you want to perform a complete format analysis?
Academicly this is the correct proceeding.

But from a practical point of view: does Noesis show up the model with its texture?
What's your final goal? To get the weighted model into Noesis?

I created an obj for blender using these faces
f 1/1/1 2/2/2 3/3/3 6/6/6
f 6/6/6 2/2/2 5/5/5 4/4/4
f 4/4/4 2/2/2 7/7/7 8/8/8
f 9/9/9 10/10/10 11/11/11 14/14/14
f 14/14/14 10/10/10 13/13/13 12/12/12
f 12/12/12 10/10/10 15/15/15 16/16/16
f 17/17/17 18/18/18 19/19/19 22/22/22
f 22/22/22 18/18/18 21/21/21 20/20/20
f 20/20/20 18/18/18 23/23/23 24/24/24
...
and the above v,vn,vts but what was displayed then did not look like a model.
(It's my first time using quads on blender 2.49 wavefront importer so maybe I missed/misunderstood something.)

[fierce]

Does Noesis use a (somehow fitting) python plugin for Ninja Storm files? (If so it would be imho easier to change this plugin for your needs.)

I have no idea where noesis stores the plugin/data to do that extraction, I did a binary search for the headers (NDP3 and XFBIN) and only found data on DLLs, not on python scripts.

To my knowledge the vertex group (numbers) are needed for skinning.

Skinning?

Seems you want to perform a complete format analysis?
Academicly this is the correct proceeding.

I think I already did a complete format analysis, at least I know what almost all of the values mean. If that's what you mean by "complete format analysis"

But from a practical point of view: does Noesis show up the model with its texture

It does, although most of the times the texture offset is wrong (Increasing it by 1 solves the problem)

What's your final goal? To get the weighted model into Noesis?

To do the exact same thing that Noesis already does with those game files, but working on all files. Like I said before, a lot of the files have wrong offsets. I'm writing a Noesis script to do that, read the xfbin file, display the models correctly, with all the textures correctly assigned. And share it with everyone here of course

I think I'm already half-way on the script. I got stumped now by not knowing how I'm going to handle the quads data. Noesis doesn't support the insertion of quad data? Do I have to convert them to tris by myself on the script?

I created an obj for blender using these faces
f 1/1/1 2/2/2 3/3/3 6/6/6
f 6/6/6 2/2/2 5/5/5 4/4/4
f 4/4/4 2/2/2 7/7/7 8/8/8
f 9/9/9 10/10/10 11/11/11 14/14/14
f 14/14/14 10/10/10 13/13/13 12/12/12
f 12/12/12 10/10/10 15/15/15 16/16/16
f 17/17/17 18/18/18 19/19/19 22/22/22
f 22/22/22 18/18/18 21/21/21 20/20/20
f 20/20/20 18/18/18 23/23/23 24/24/24
...
and the above v,vn,vts but what was displayed then did not look like a model.
(It's my first time using quads on blender 2.49 wavefront importer so maybe I missed/misunderstood something.)

If I'm not mistaken those faces are from a different file than the one I pasted the sample v,vn,vts. My bad, sorry.

[shakotay2]

I have no idea where noesis stores the plugin/data to do that extraction, I did a binary search for the headers (NDP3 and XFBIN) and only found data on DLLs, not on python scripts.

Ok, then it's indeed up to you to write a plugin.

Skinning?

The data that "tells" the game engine which vertices (i.e. vertex groups) to move when an NPC raises his arm for example. (As far as I understood you can create vertex groups bei weighted painting for example.)

[...]at least I know what almost all of the values mean. If that's what you mean by "complete format analysis"

More or less. Depends on your final goal. As on your following answer I guess your analysis is complete.

To do the exact same thing that Noesis already does with those game files, but working on all files.

I'm not familiar with Noesis. If it is capable of showing up the skeleton I would take a glance. Meanwhile I'm using blender and Gmax. With Gmax the vertex painting seems to be a little bit easier but I'm just at the beginning (since 1 or two years or so ).

And share it with everyone here of course

That would be great.

Noesis doesn't support the insertion of quad data? Do I have to convert them to tris by myself on the script?

As I said I'm not familiar with Noesis. There is a guy in this forum, finale00, who could answer this, I'm sure.

If I'm not mistaken those faces are from a different file than the one I pasted the sample v,vn,vts. My bad, sorry.

Well, I took the faces from si00bobj006_01_1.xfbin, address 0x124 and the v,vn,vts from
0x524. But I chose an estimated verts count of 272; maybe I should increase it. edit: nope

but there are faces with 5 vertices -> quints???

[fierce]

Well, I took the faces from si00bobj006_01_1.xfbin, address 0x124 and the v,vn,vts from
0x524. But I chose an estimated verts count of 272; maybe I should increase it. edit: nope

but there are faces with 5 vertices -> quints???

The mesh looks somewhat like that, but you have some messed up verts/faces. All the "separated objects" on that mesh have 8 verts and 6 tris. I dont think there are any faces with 5 vertices. On what address did you see that?

[shakotay2]

I dont think there are any faces with 5 vertices. On what address did you see that?

0x30E and 0x324 for example:

Code: Select all

Offset  0  1  2  3  4  5  6  7   8  9  A  B  C  D  E  F

000300 00 7F FF FF 00 80 00 81  00 82 00 87 FF FF 00 85  
000310 00 86 00 81 00 84 00 83  FF FF 00 88 00 89 00 8A  
000320 00 8F FF FF 00 8D 00 8E  00 89 00 8C 00 8B FF FF  
000330 00 90 00 91 00 92 00 97  FF FF 00 95 00 96 00 91 

[fierce]

I dont think there are any faces with 5 vertices. On what address did you see that?

0x30E and 0x324 for example:

Code: Select all

Offset  0  1  2  3  4  5  6  7   8  9  A  B  C  D  E  F

000300 00 7F FF FF 00 80 00 81  00 82 00 87 FF FF 00 85  
000310 00 86 00 81 00 84 00 83  FF FF 00 88 00 89 00 8A  
000320 00 8F FF FF 00 8D 00 8E  00 89 00 8C 00 8B FF FF  
000330 00 90 00 91 00 92 00 97  FF FF 00 95 00 96 00 91 

I tried to parse the values without ignoring "FFFF", at least not treating it as a separator and the values match. But they dont seem to make sense that way.

Also there seems to be faces with only 3 values.
There seem to be:
7 quints and 7 tris

The VERT/FACE group on that file indicates a number that I'm sure its related to faces: 509
But on the header it says that the face data size is 1024 but I there only seems to be 1018 bytes of face data
(what's left is 0000 0000 on the beginning and 0000 0000 on the end that doesn't seem to fit the face data)
And 1018 makes much more sense than 1024, because 509x2 = 1018.

After exported by Noesis the .obj file has 204 faces.
If 7 quints have 3 faces (tris) each = 21
There are also 7 tris on the faces = 7
And there are 88 quads = 176
21+7+176 = 204.
So I think it kind of makes sense that it has 7 quints, 7 tris and 88 quads.
But why is it formed like that? Is it normal for mesh to be mixed with quints tris and quads?

I wish I had more experience with this.
What is a quint? A pentagon? I'm not very good with geometry either and cant seem to find any info on google.
Also how do quints transform into quads or tris? So I can check if the data matches.

Thank you.

The parsed data (considering FFFF as a separator) just to help:

1: - 1 - 2 - 3 - 6 - FFFF
2: - 6 - 2 - 5 - 4 - FFFF
3: - 4 - 2 - 7 - 8 - FFFF
4: - 9 - 10 - 11 - 14 - FFFF
5: - 14 - 10 - 13 - 12 - FFFF
6: - 12 - 10 - 15 - 16 - FFFF
7: - 17 - 18 - 19 - 22 - FFFF
8: - 22 - 18 - 21 - 20 - FFFF
9: - 20 - 18 - 23 - 24 - FFFF
10: - 25 - 26 - 27 - 30 - FFFF
11: - 30 - 26 - 29 - 28 - FFFF
12: - 28 - 26 - 31 - 32 - FFFF
13: - 33 - 34 - 35 - 38 - FFFF
14: - 38 - 34 - 37 - 36 - FFFF
15: - 36 - 34 - 39 - 40 - FFFF
16: - 41 - 42 - 43 - 46 - FFFF
17: - 46 - 42 - 45 - 44 - FFFF
18: - 44 - 42 - 47 - 48 - FFFF
19: - 49 - 50 - 51 - 54 - FFFF
20: - 54 - 50 - 53 - 52 - FFFF
21: - 52 - 50 - 55 - 56 - FFFF
22: - 57 - 58 - 59 - 62 - FFFF
23: - 62 - 58 - 61 - 60 - FFFF
24: - 60 - 58 - 63 - 64 - FFFF
25: - 65 - 66 - 67 - 70 - FFFF
26: - 70 - 66 - 69 - 68 - FFFF
27: - 68 - 66 - 71 - 72 - FFFF
28: - 73 - 74 - 75 - 78 - FFFF
29: - 78 - 74 - 77 - 76 - FFFF
30: - 76 - 74 - 79 - 80 - FFFF
31: - 81 - 82 - 83 - 86 - FFFF
32: - 86 - 82 - 85 - 84 - FFFF
33: - 84 - 82 - 87 - 88 - FFFF
34: - 89 - 90 - 91 - 94 - FFFF
35: - 94 - 90 - 93 - 92 - FFFF
36: - 92 - 90 - 95 - 96 - FFFF
37: - 97 - 98 - 99 - 102 - FFFF
38: - 102 - 98 - 101 - 100 - FFFF
39: - 100 - 98 - 103 - 104 - FFFF
40: - 105 - 106 - 107 - 110 - FFFF
41: - 110 - 106 - 109 - 108 - FFFF
42: - 108 - 106 - 111 - 112 - FFFF
43: - 113 - 114 - 115 - 118 - FFFF
44: - 118 - 114 - 117 - 116 - FFFF
45: - 116 - 114 - 119 - 120 - FFFF
46: - 121 - 122 - 123 - 126 - FFFF
47: - 126 - 122 - 125 - 124 - FFFF
48: - 124 - 122 - 127 - 128 - FFFF
49: - 129 - 130 - 131 - 136 - FFFF
50: - 134 - 135 - 130 - 133 - 132 - FFFF
51: - 137 - 138 - 139 - 144 - FFFF
52: - 142 - 143 - 138 - 141 - 140 - FFFF
53: - 145 - 146 - 147 - 152 - FFFF
54: - 150 - 151 - 146 - 149 - 148 - FFFF
55: - 153 - 154 - 155 - 160 - FFFF
56: - 158 - 159 - 154 - 157 - 156 - FFFF
57: - 161 - 162 - 163 - 168 - FFFF
58: - 166 - 167 - 162 - 165 - 164 - FFFF
59: - 169 - 170 - 171 - 176 - FFFF
60: - 174 - 175 - 170 - 173 - 172 - FFFF
61: - 177 - 178 - 179 - 184 - FFFF
62: - 182 - 183 - 178 - 181 - 180 - FFFF
63: - 185 - 186 - 187 - 190 - FFFF
64: - 190 - 186 - 189 - 188 - FFFF
65: - 188 - 186 - 191 - 192 - FFFF
66: - 193 - 194 - 195 - 198 - FFFF
67: - 198 - 194 - 197 - 196 - FFFF
68: - 196 - 194 - 199 - 200 - FFFF
69: - 201 - 202 - 203 - 206 - FFFF
70: - 206 - 202 - 205 - 204 - FFFF
71: - 204 - 202 - 207 - 208 - FFFF
72: - 209 - 210 - 211 - 214 - FFFF
73: - 214 - 210 - 213 - 212 - FFFF
74: - 212 - 210 - 215 - 216 - FFFF
75: - 217 - 218 - 219 - 222 - FFFF
76: - 222 - 218 - 221 - 220 - FFFF
77: - 220 - 218 - 223 - 224 - FFFF
78: - 225 - 226 - 227 - 230 - FFFF
79: - 230 - 226 - 229 - 228 - FFFF
80: - 228 - 226 - 231 - 232 - FFFF
81: - 233 - 234 - 235 - 238 - FFFF
82: - 238 - 234 - 237 - 236 - FFFF
83: - 236 - 234 - 239 - 240 - FFFF
84: - 241 - 242 - 243 - 246 - FFFF
85: - 246 - 242 - 245 - 244 - FFFF
86: - 244 - 242 - 247 - 248 - FFFF
87: - 249 - 250 - 251 - 254 - FFFF
88: - 254 - 250 - 253 - 252 - FFFF
89: - 252 - 250 - 255 - 256 - FFFF
90: - 257 - 258 - 259 - 262 - FFFF
91: - 262 - 258 - 261 - 260 - FFFF
92: - 260 - 258 - 263 - 264 - FFFF
93: - 265 - 266 - 267 - 270 - FFFF
94: - 270 - 266 - 269 - 268 - FFFF
95: - 268 - 266 - 271 - 272 - FFFF
96: - 130 - 132 - 136 - FFFF
97: - 138 - 140 - 144 - FFFF
98: - 146 - 148 - 152 - FFFF
99: - 154 - 156 - 160 - FFFF
100: - 162 - 164 - 168 - FFFF
101: - 170 - 172 - 176 - FFFF
102: - 178 - 180 - 184 -

[fierce]

I just found on one file that I have different sizes of face entries:
3, 4, 5, 7, 9, 13, 14
Using my script that treats "FFFF" as a separator I get:
(for the first 2 groups)

Face Offset: 688
Face Size: 191
[1, 2, 0, 3]
[9, 8, 5, 6, 4, 7, 17, 16, 15, 14, 13, 10, 12, 11]
[2, 18, 3, 19]
[21, 22, 20, 23, 38]
[29, 28, 25, 26, 24, 27, 37, 36, 35, 34, 33, 30, 32, 31]
[45, 44, 41, 42, 40, 43, 53, 52, 51, 50, 49, 46, 48, 47]
[96, 97, 94, 95, 92, 93, 90, 91, 89, 86, 88, 87, 98, 99]
[101, 102, 100, 103, 118]
[109, 108, 105, 106, 104, 107, 117, 116, 115, 114, 113, 110, 112, 111]
[62, 61, 55, 56, 54, 57, 58, 64, 65]
[61, 63, 56, 57]
[54, 58, 59, 65, 60, 66, 62]
[59, 60, 54, 55]
[70, 67, 69, 68, 79, 80, 85, 84, 83, 75, 73, 74, 72]
[74, 75, 76, 77, 78, 68, 67]
[75, 84, 77, 80, 68]
[70, 69, 71, 81, 82, 85, 83]
[70, 71, 72, 73]
[69, 79, 81, 85]
[71, 82, 73, 83]
[38, 39, 20]
[57, 63, 64]
[62, 55, 60]
[118, 119, 100]
Face Offset: 1070
Face Size: 169
[1, 2, 0, 3]
[5, 6, 4, 7]
[9, 10, 8, 11]
[13, 14, 12, 15]
[17, 18, 16, 19]
[21, 22, 20, 23]
[25, 26, 24, 27]
[29, 30, 28, 31]
[33, 34, 32, 35]
[37, 38, 36, 39]
[41, 42, 40, 43]
[45, 46, 44, 47]
[49, 50, 48, 51]
[53, 54, 52, 55]
[57, 58, 56, 59]
[61, 62, 60, 63]
[65, 66, 64, 67]
[69, 70, 68, 71]
[73, 74, 72, 75]
[77, 78, 76, 79]
[81, 82, 80, 83]
[85, 86, 84, 87]
[89, 90, 88, 91]
[93, 94, 92, 95]
[97, 98, 96, 99]
[101, 102, 100, 103]
[105, 106, 104, 107]
[109, 110, 108, 111]
[113, 114, 112, 115]
[117, 118, 116, 119]
[121, 122, 120, 123]
[125, 126, 124, 127]
[129, 130, 128, 131]
[133, 134, 132, 135]

Could they be polygons with different sizes or something like that?