error

Update 2016-08-16:

Hmm... I tried Python 2.7 but the problem remains.

Hi Michael,

First of all - thank you for the excellent package!

I think that it is because I am using Python 3.5 (Anaconda Python 3.5 64 bit in 64 bit Windows 10), I have this error (potentially due to the relative import mechanism).

When I do the following code snippet:

import numpy as np
import trimesh

if __name__ == '__main__':
    mesh = trimesh.load_mesh(test.stl')
    section = mesh.section(plane_origin=[0, 0, 0],
                           plane_normal=[0, 0, 1])

I see the following error info:

runfile('C:/ProjectFolder/MaterialFEA/geometry/Export/test.py', wdir='C:/ProjectFolder/MaterialFEA/geometry/Export')
Traceback (most recent call last):

  File "<ipython-input-14-a216ee98491b>", line 1, in <module>
    runfile('C:/ProjectFolder/MaterialFEA/geometry/Export/test.py', wdir='C:/ProjectFolder/MaterialFEA/geometry/Export')

  File "C:\Anaconda3\lib\site-packages\spyderlib\widgets\externalshell\sitecustomize.py", line 714, in runfile
    execfile(filename, namespace)

  File "C:\Anaconda3\lib\site-packages\spyderlib\widgets\externalshell\sitecustomize.py", line 89, in execfile
    exec(compile(f.read(), filename, 'exec'), namespace)

  File "C:/ProjectFolder/MaterialFEA/geometry/Export/test.py", line 15, in <module>
    plane_normal=[0, 0, 1])

  File "C:\Anaconda3\lib\site-packages\trimesh\base.py", line 945, in section
    path = load_path(lines)

NameError: name 'load_path' is not defined

Do you think it is the python 2 vs 3 compatibility issue due to the relative imports? Any ideas on whether there is a quick fix available?

Thank you!

Shawn

PS: The stl file is just a simple cube for testing and I don't think that is the issue.

I'm using the mesh.section function to detect intersections. The problem is that a lot of our data is incorrect geometry (missing faces), and the library discards those segments completely. Having that data is essential when dealing with bad files.

Hi,

Very simple piece of code leads to the following error warning, although it does not break the code execution. I am not 100% sure it is trimesh related, however, any clarification would be helpful.

The issue is following.
I load a mesh from an 'obj' file and execute ray.inersects_any function by using a single ray origin and unit vector. Since there is no single intersection the function returns 'False'. In the next line I load again the same 'obj' file (assigning a different variable name) and execute again a ray.inersects_any function. The following three lines of error warnings appears before the output of the function (which is again 'False'):

ERROR CAUGHT IN EMBREE
ERROR: Invalid operation
ERROR MESSAGE: b'already initialized'

The code is:

import trimesh

ray_origin = [[5, -0.001, 2.5]]
vector = [[0, -1, 0]]

mesh = trimesh.load_mesh('./test_mesh.obj')
intersect = mesh.ray.intersects_any(ray_origin, vector)
print(intersect)

mesh1 = trimesh.load_mesh('./test_mesh.obj')
intersect = mesh1.ray.intersects_any(ray_origin, vector)
print(intersect)

While the 'test_mesh.obj' is:

v 0 0 5
v 0 0 0
v 10 0 5
v 10 0 0
v 10 10 5
v 10 10 0
v 0 10 5
v 0 10 0
f 1 2 3
f 3 2 4
f 3 4 5
f 5 4 6
f 5 6 7
f 7 6 8
f 7 8 1
f 1 8 2

Best regards

Hi,

Is it possible to update a mesh, specifically its vertex colours, while it's being shown with mesh.show()?
I found a mesh.visual.update_vertices(mask=...) method but I don't think it does what I'm looking for.

Specifically, if I wanted to add a slider to the viewer that changes vertex colour and position values of the mesh, and the mesh should update in real-time in the viewer. Do you think that's feasible with trimesh and pyglets? Would I probably best start by copying viewer.py?

Any advice welcome - I'm looking for a simple tool that would allow me to achieve this. I've also been thinking about using Blender's Python interface for this.

Hi Mike,

(I assume this would be a simple question :)

Would there be any simple way to decimate a mesh? Say, simplify the geometry from 10,000 faces to 1,000 faces?

Thanks!

Shawn

file: tirangle.py
function: mass_properties
problem: Missing value inertia[1,0]

inertia = np.zeros((3,3))
inertia[0,0] = integrated[5] + integrated[6] - (volume * (center_mass[[1,2]]**2).sum())
inertia[1,1] = integrated[4] + integrated[6] - (volume * (center_mass[[0,2]]**2).sum())
inertia[2,2] = integrated[4] + integrated[5] - (volume * (center_mass[[0,1]]**2).sum())
inertia[0,1] = (integrated[7] - (volume * np.product(center_mass[[0,1]])))
inertia[1,2] = (integrated[8] - (volume * np.product(center_mass[[1,2]])))
inertia[0,2] = (integrated[9] - (volume * np.product(center_mass[[0,2]])))
inertia[2,0] = inertia[0,2]
inertia[2,1] = inertia[1,2] 

-----------------

Add the following line which is missing

-----------------

inertia[1,0] = inertia[0,1]

Hi Michael,

yesterday I came across your code because I have to merge a lot of points of a surface triangulation. After some testing I came across a bug in the "merge_vertices()" function of your code.
Here is a minimal example:

import numpy as np
import trimesh
nds2 = np.array([[ 1. , 1.60994057, 2.07798185],
[ 1. , 0.60994057, 2.07798184],
[ 1. , 0.60994057, 2.07798185]])
ele2 = np.array([[0,1,2]])
mesh = trimesh.Trimesh(vertices=nds2, faces=ele2,process=True)
mesh.merge_vertices()
print mesh.vertices
print np.linalg.norm(mesh.vertices[1]-mesh.vertices[2])

Best,
Konrad

I have sliced a 3d mesh with yz planes into several 2d sections (i.e. Path2D objects). Now, I'd like to create Shapely polygons from these 2d sections. For doing so, I need a list of points as XY coords. However, if I use Path2D.vertices, I get a weird polygon because the vertices are un-ordered. Is there any way to get an ordered list of the vertices composing a Path2D object?

Obviously I cannot use the python .sort() method.

In complete analogy to Trimesh.faces_sparse from #19, I have a use case for a cached sparse matrix encoding the mapping vertex -> edges:

edges = self.face_adjacency_edges
row = edges.reshape(-1)
col = np.tile(np.arange(len(edges)).reshape(-1, 1), (1, 2)).reshape(-1)
self.edges_sparse = sp.coo_matrix(
    (np.ones(len(col), dtype=np.bool), (row, col)),
    shape=(vertex_count, len(edges)), dtype=np.bool)

Then

G = np.zeros((vertex_count, 3))
for edge, f in zip(self.face_adjacency_edges, F):
    G[edge] += f

is equivalent to:

G = self.edges_sparse.dot(F)

Hi! Thanks for the great library!

Let's consider the following code:

    paths = mesh.outline().paths
    for p in paths:
        entities = mesh.outline().entities[p]
        ee = []
        for e in entities:
            ee.extend(e.points)
        print(ee)

Which is just extracting the list of vertex indexes from each outline. In my case the following result is obtained:

[0, 19, 0, 69, 168, 69, 192, 168, 192, 124, 138, 124, 138, 8, 8, 98, 98, 76, 197, 76, 197, 112, 21, 112, 59, 21, 59, 177, 9, 177, 108, 9, 22, 108, 22, 157, 126, 157, 194, 126, 6, 194, 6, 73, 213, 73, 139, 213, 139, 89, 89, 180, 39, 180, 18, 39, 93, 18, 93, 55, 55, 20, 101, 20, 129, 101, 129, 64, 64, 58, 42, 58, 42, 170, 170, 154, 209, 154, 128, 209, 128, 153, 46, 153, 46, 37, 71, 37, 71, 164, 164, 115, 115, 48, 121, 48, 121, 120, 30, 120, 150, 30, 150, 137, 137, 176, 77, 176, 188, 77, 17, 188, 17, 191, 191, 195, 195, 106, 205, 106, 158, 205, 158, 181, 65, 181, 65, 19]
[12, 107, 12, 140, 140, 97, 97, 27, 160, 27, 160, 60, 60, 100, 100, 11, 63, 11, 63, 52, 29, 52, 29, 184, 165, 184, 3, 165, 3, 56, 167, 56, 204, 167, 204, 207, 207, 13, 136, 13, 136, 196, 62, 196, 62, 202, 202, 151, 50, 151, 50, 155, 211, 155, 211, 2, 4, 2, 4, 78, 61, 78, 61, 198, 198, 161, 70, 161, 70, 199, 47, 199, 32, 47, 32, 24, 44, 24, 44, 51, 148, 51, 148, 15, 86, 15, 86, 111, 111, 118, 45, 118, 45, 28, 28, 144, 49, 144, 49, 85, 43, 85, 43, 5, 34, 5, 34, 212, 212, 105, 178, 105, 178, 125, 125, 94, 94, 147, 16, 147, 53, 16, 53, 75, 25, 75, 25, 103, 103, 182, 182, 166, 145, 166, 169, 145, 102, 169, 102, 14, 206, 14, 206, 67, 67, 119, 1, 119, 1, 107]
[146, 92, 189, 92, 189, 23, 132, 23, 90, 132, 90, 82, 82, 81, 81, 80, 79, 80, 83, 79, 91, 83, 91, 133, 133, 26, 186, 26, 186, 190, 33, 190, 33, 152, 117, 152, 117, 193, 40, 193, 57, 40, 185, 57, 185, 95, 95, 174, 141, 174, 74, 141, 74, 179, 135, 179, 172, 135, 172, 163, 163, 162, 130, 162, 130, 66, 66, 84, 84, 201, 68, 201, 116, 68, 116, 131, 131, 54, 54, 31, 200, 31, 200, 88, 159, 88, 159, 203, 41, 203, 41, 114, 114, 143, 142, 143, 113, 142, 38, 113, 38, 183, 183, 99, 171, 99, 208, 171, 208, 110, 173, 110, 87, 173, 87, 96, 72, 96, 210, 72, 210, 109, 10, 109, 10, 156, 134, 156, 134, 127, 123, 127, 123, 122, 104, 122, 104, 35, 175, 35, 175, 149, 187, 149, 187, 36, 7, 36, 7, 146]

That is, 3 outlines. In the former two cases, I was expecting that the last point of each edge was the first point of the next one ([19, 0, 0, 69, 69, 168, 168, 192, 192, ...]).

Anyway, assuming that such weird organization is right, the last outline has something strange at the end... i.e. concatenating the points list with itself you would ge something like [..., 36, 7, 146, 146, 92, 189, ...]

Am I doing something wrong? Is it the expected behavior?
I know that paths is only returning the list of entities from the mesh.outline(), and therefore which its orientation. However, looking at the code of closed_paths() function, it seems you are caring about the orientation

Thanks!
Pepe

Hi!

When I display a mesh with mesh.show(), rotating it with the left mouse works, until I press 'z' to reset the view - then, it only rotates around the z-axis (roll-angle), and not around the other two axes anymore. It's 100% reproducible.

This is on Win 10 in a virtualenv with python-3.6.1 (64-bit), numpy+mkl, scipy and Shapely from here, and pip install trimesh[easy].

The following code segfaults for me after 15-20 iterations:

import trimesh
from copy import deepcopy

mesh = trimesh.primitives.Sphere()
mesh.triangles_tree()

print type(mesh._cache.cache['_triangles_tree'])

for i in range(0,30):
    print i
    deepcopy(mesh)

print "done"

(This was distilled for a much larger failing example that was originally calling mesh.copy(), which is implemented with deepcopy()).

Because mesh.copy() breaks, so do all of the functions that depend on it for their implementation.

I suspect the problem is the rtree.index.Index object stored in the mesh cache in the slot '_triangles_tree' -- it doesn't look like it supports deepcopy.

Consider: clearing the cache before copying (seems to work in a simple example), OR, to implement copy / deepcopy in the util.Cache object to do the right thing.

Looks like the STLs generated are only binary, without any option for ASCII.

The pip install trimesh worked fine, but when I attempted to import the module (import trimesh) it failed since shapely was not found.

Info: windows 10, WinPython 3.5.1.2, Python 3.5, 64bit

how do you install libspatialindex-dev on a mac? I only saw windows builds here
https://libspatialindex.github.io/
In order to run convexhull.py, I need rtree which in turn needs libspatialindex-dev

Hi,

My apologies if this is not the right place for this question. Could you please tell me how to generate the documentation for the trimesh library? I have not yet found the readme that explains how it is done! Many thanks.

Adrian Butscher

How would one go by getting an the 1 ring neighbhourhood for all vertices? I can see trimesh has functions for adjacent faces, but not for adjacent vertices.

@mikedh I am currently using a virtual environment for Python 3.3.2 with Anaconda on windows 7. How could I set up trimesh to be compatible with Python 3.3.2? https://docs.continuum.io/anaconda/old-pkg-lists/2.3.0/py33

I installed trimesh via pip but I wasn't able to get it to work with my Anaconda python 3.3 virtual environment

It seems the MacOS implementation of the window doesn't render with visible=false. This affects both the following ways of generating an image:

mesh.scene().save_image('image.png')
and
mesh.show(save_image='image.png')

As a work around, could you please add the option to set the visible flag to True even when using save_image?

Otherwise the workaround is to duplicate much of SceneViewer() to force visible to True

Hello there! I have recently started using your package and find it very appealing. Thank you for sharing your work!

It seems that geometry.mean_vertex_normals loses information about the normals of some incident faces, i.e. when the list of indices faces[:,i] contains repetitions. Instead, to create a field G on vertices by accumulating a field F on incident faces, I use:

G = np.zeros([len(self.vertices)] + list(F.shape[1:]))
for i, face in enumerate(self.faces):
    G[face] += F[i]

The mapping vertex -> incident faces could be cached as a sparse matrix, but for tasks similar to this, the necessary lookups probably wouldn't improve performance.

I've successfully loaded a little mesh and checked it for watertightness, etc., and I'm trying to 'show' it in a pyglet window using some code I cribbed from trimesh's examples. I know my pyglet / pygame / opengl installation is (at least somewhat) correct because I can run their samples standalone, e.g., THIS ONE. The trimesh code is shown below.

import numpy as np
import trimesh

mesh = trimesh.load_mesh('../models/j39binary.fbx')
mesh.is_watertight # ~~> True
mesh.euler_number # ~~> 2
np.divide(mesh.volume, mesh.convex_hull.volume) # ~~> works as expected
mesh.vertices -= mesh.center_mass # ditto
mesh.moment_inertia # ~~> results look reasonable
mesh.split() # ~~> ok
for facet in mesh.facets:
    mesh.visual.face_colors[facet] = trimesh.visual.random_color()

mesh.show() # ~~> throws inside here
# omitting the rest

The stack trace is below. The symptom is that _nswindow in the SceneViewer is None. I've stepped through a ton of pyglet code. The problem seems to be that a correctly created _nswindow is thrown away on line 209 of gl/__init__py with a call of app.windows.remove. Later on, SceneViewer tries to use _nswindow, but it's gone. This appears to be due to some multithreading issues or some tricks for creating windows in the face of circular imports, it's hard to say from the code I stepped through.

I just don't know enough about pyglet and its Cocoa implementation to figure out a workaround, so I thought I'd bring this to you. Simple things like trying to create a new window right inside trimesh's viewer.py and assigning it to _nswindow caused more problems and the exercise is becoming a classic rabbit hole. I'm sure it's something simple, but I'm not sure it's my problem because I copy-pasted the calling code from trimesh samples. Also, I am not the only one with the problem: please see THIS STACKOVERFLOW QUESTION. I'd be grateful for any advice.

Traceback (most recent call last):
  File "/Applications/PyCharm CE.app/Contents/helpers/pydev/pydevd.py", line 1591, in <module>
    globals = debugger.run(setup['file'], None, None, is_module)
  File "/Applications/PyCharm CE.app/Contents/helpers/pydev/pydevd.py", line 1018, in run
    pydev_imports.execfile(file, globals, locals)  # execute the script
  File "/Users/bbeckman/Documents/trimesh/examples/debugio.py", line 43, in <module>
    mesh.show()
  File "/Users/bbeckman/Documents/trimesh/trimesh/base.py", line 1373, in show
    scene.show(**kwargs)
  File "/Users/bbeckman/Documents/trimesh/trimesh/scene/scene.py", line 388, in show
    viewer()
  File "/Users/bbeckman/Documents/trimesh/trimesh/scene/scene.py", line 385, in viewer
    SceneViewer(self, **kwargs)
  File "/Users/bbeckman/Documents/trimesh/trimesh/scene/viewer.py", line 31, in __init__
    self.reset_view(flags=flags)
  File "/Users/bbeckman/Documents/trimesh/trimesh/scene/viewer.py", line 129, in reset_view
    if (self.width is not None and
  File "/Users/bbeckman/anaconda/envs/rsimcon-pyassimp-minimal/lib/python2.7/site-packages/pyglet/window/__init__.py", line 850, in <lambda>
    width = property(lambda self: self.get_size()[0],
  File "/Users/bbeckman/anaconda/envs/rsimcon-pyassimp-minimal/lib/python2.7/site-packages/pyglet/window/cocoa/__init__.py", line 398, in get_size
    window_frame = self._nswindow.frame()
AttributeError: 'NoneType' object has no attribute 'frame'

Hi,

I have an edge case where voxelising a mesh is incorrectly calculating the number of voxels required due to rounding error. The issue is on line 68 in voxel.py, specifically:

raw_shape = np.ptp(bounds/pitch, axis=0).astype(int)

One of the axes is coming out at 14.99999999999999822364, which is converted to 14 as an integer. This later results in out of range indexing in run_to_raw() (voxel.py line 43).

The following change is one viable solution:
raw_shape = (np.ptp(bounds/pitch, axis=0) + (0.5, 0.5, 0.5)).astype(int)

Thanks.

Hi, I have the following error when loading a mesh.

I am trying to load this off file by

import trimesh
mesh = trimesh.load_mesh('chair.off', 'off', process=False)

The header in the off file is

OFF
37580 74488 0

But then the loaded mesh has only 74487 faces. Do you have any idea what's going on here or do I miss anything?
Thank you very much.

Hi

Older (much older) version of trimesh worked fine for me.
After updating (now 2.2.11), it does not manage to read the PLY file, I get

File "/usr/lib/python2.7/site-packages/trimesh/io/ply.py", line 131, in elements_to_kwargs
faces = elements['face']['data']['vertex_indices']['f1']
ValueError: no field of name vertex_indices

Any ideas on what is the issue in the updated version?

If you load dxf file that has objects on different layers.
When loaded, do you know what layer it came from?

Hi, can you support models in ascii ply format?

Here is what I got when trying to load a ply file.

'''
51 if 'ascii' in encoding:
---> 52 raise ValueError('ASCII PLY not supported!')
53
54 endian = ['<', '>']['big' in encoding]
'''

Hi Michael,

Quick question - would the solid name be stored somewhere after the STL file got loaded?

Thanks!

Shawn

Hi Michael,
my name is Johann from Germany and I did found trimesh very interesting.
The main purpose is cutting a mesh with a plane.
Therefore I did tested section.py and it works well well also if I did cutted featuretype.STL also in the y or x direction changing the code
y_extents = mesh.bounds[:,1]

# slice every .125 model units (eg, inches)
y_levels  = np.arange(*y_extents, step=0.25)
print (y_levels)
# create an array to hold the section objects
sections  = [None] * len(y_levels)

for i, y in enumerate(y_levels):
    print (i, y)
    # this will return a Path3D object, each of which will 
    # have curves in 3D space
    sections[i] = mesh.section(plane_origin = [0,y,0],                                                
                               plane_normal = [0,1,0])

Everything is working fine until I did used a different mesh-file skinabdomen.STL ( body surface generated from dicom CT files)
The script is working for this file for the Z cut as described below:

mesh = trimesh.load_mesh('/home/medinex/trimesh-master/models/skinabdomen.STL')
#mesh = trimesh.load_mesh('/home/medinex/trimesh-master/models/featuretype.STL')
mesh.show()
print (mesh.bounding_box_oriented.primitive.extents)
# we're going to slice the mesh into evenly spaced chunks along z
# this takes the (2,3) bounding box and slices it into [minz, maxz]
z_extents = mesh.bounds[:,2]

# slice every .125 model units (eg, inches)
z_levels  = np.arange(*z_extents, step=10)
print (z_levels)
# create an array to hold the section objects
sections  = [None] * len(z_levels)

for i, z in enumerate(z_levels):
    print (i,z)
    # this will return a Path3D object, each of which will 
    # have curves in 3D space
    sections[i] = mesh.section(plane_origin = [0,0,z],
                               plane_normal = [-0.001,0,1])

In case of Y direction cut
mesh = trimesh.load_mesh('/home/medinex/trimesh-master/models/skinabdomen.STL')
#mesh = trimesh.load_mesh('/home/medinex/trimesh-master/models/featuretype.STL')
mesh.show()
print (mesh.bounding_box_oriented.primitive.extents)
# we're going to slice the mesh into evenly spaced chunks along z
# this takes the (2,3) bounding box and slices it into [minz, maxz]
y_extents = mesh.bounds[:,1]

# slice every .125 model units (eg, inches)
y_levels  = np.arange(*y_extents, step=30)
print (y_levels)
# create an array to hold the section objects
sections = [None]*len(y_levels)
print (len(y_levels))
for j, y in enumerate(y_levels):
    print (j,y)
    # this will return a Path3D object, each of which will 
    # have curves in 3D space
    sections[j] = mesh.section(plane_origin = [0,y,0],
                               plane_normal = [0,1,0])

The error is displayed:

Traceback (most recent call last):
File "/home/medinex/trimesh-master/trimesh/section-skinabdomeny.py", line 43, in
plane_normal = [0,1,0])
File "/usr/local/lib/python2.7/dist-packages/trimesh/base.py", line 998, in section
raise ValueError('Specified plane doesn't intersect mesh!')
ValueError: Specified plane doesn't intersect mesh!

I looked to the base.py, trying to detect the error and I have seen that in your original code and your original line below it is : from .io.load import load_path
Should it maybe be : from .path.io.load import load_path because load path is only there.

your code :
from .io.load import load_path
lines = intersections.mesh_plane(mesh = self,
plane_normal = plane_normal,
plane_origin = plane_origin)
if len(lines) == 0:
raise ValueError('Specified plane doesn't intersect mesh!')
path = load_path(lines)
return path
trying to detect the error.

Please let me know if maybe you see a reason !!
The y direction cut is working without problem for the featuretype.STL file.

Do you have maybe a more detailed description how to use intersections.py from trimesh ?
Thank you in advance
Best regards
Johann

I'm exercising trimesh on some FBX models. I backed off to Pyassimp after having bad luck with Cyassimp (unexplained IO failures in the C code), looking for ways to make quicker progress (I may one day get back to Cyassimp, but I need to move quickly now). I am having good luck with Pyassimp, but I noticed some differences in the shape of results returned to trimesh from Pyassimp when Pyassimp imports FBX models and when Pyassimp imports STL models. I worked around the differences at the trimesh level, but my workarounds aren't sufficiently well thought-out (by a very long margin) to submit as a pull request. In fact, this entire issue may be better addressed as an assimp issue, and I will be digging in there as time allows. In the meantime, I thought I'd bring it to your attention. Perhaps the following code snippet from near line 120 in load.py is self-explanatory. This is what I did to get trimesh to work.

    # make sure we keep passed kwargs to loader
    # but also make sure loader keys override passed keys
    loader_keys = mesh_loaders[file_type](file_obj, file_type)
    # [bbeckman: introduce the variable above to track in the debugger.
    # Apparently, the fbx loader produces a list of dictionaries, whereas the
    # STL loader, used in all the examples in trimesh, produces a dictionary.
    # That trips up kwargs.update. We'll just work around it for our cases.]
    if isinstance(loader_keys, list):
        kwargs.update(loader_keys[0])
    else:
        kwargs.update(loader_keys)

    # [bbeckman: The fbx loader returns a vertex_colors key whose value is
    # the empty list, []. That apparently causes some downstream methods to
    # silently fail. The failure occurs in the ColorVisuals call at the end
    # of visual.create_visual, which is called on line 105 of base.py. We'll
    # work around it for our cases.]
    if 'vertex_colors' in kwargs and \
            not kwargs["vertex_colors"]: # idiom for empty
        kwargs.pop("vertex_colors")

with the convexify example, i get an error message stating that "np float object does not support show method."

I'll post the code and the exact error message later. I am using a different stl file than box.stl

convex_combined = np.sum(meshes_convex)
convex_combined.show()

Hi Mike,

I came across your Trimesh library for Python after trialling several other tools for generating signed distance functions from STL files. I’ve found yours to be by far the best for what I need, but I have a simple question that I couldn’t quite figure out from looking at the source:

Essentially what I want to do is to fill the bounding box with a 3D rectangular grid of points, and compute the SDF from these. I am able to generate points and calculate the SDF with no problems, but I don’t know how to get the coordinates of the corners of the bounding box. If I use:

box = mesh.bounding_box_oriented.primitive.extents

I get the half-lengths of each side of the bounding box, but if I use these to plot the 8 corners they form a box the same size and shape as the bounding box, but seemingly randomly oriented in space. How can I correctly get the coordinates of the corners so I can construct my meshgrid within the bounding box?
I see that in bounds.py there is a function corners that seems to do what I want, but I don't know where to get the bounds argument from.

Thank you for your time,
Joel

Hi, I have a model consists of a list of 3D mesh shapes. I'm trying to find all the pairs of objects that are overlapping or touching.
The way I'm using trimesh now is to choose one mesh, traverse all the other meshes, use rtree find potential intersecting triangles, then do point contain tests to check if they are overlapping. There are three problems:

1. very slow
2. sometime the mesh has defect - not water tight, then the point contain test may fail
3. the objects touching test is not reliable

Do you have any advise on it?

Thanks!

Works fine for most slices, but crashes if it hits one with an odd number of intersections.

File "/usr/share/miniconda/lib/python2.7/site-packages/trimesh/intersections.py", line 37, in mesh_plane_intersection
return intersections.reshape(-1,2,3)
ValueError: total size of new array must be unchanged

Hi Mike,

A quick question - I noticed that the backends of blender or scad are used for boolean operations. While it is trivial to install them in Linux, are you aware that whether there would be a way to get them working in windows?

Thanks!

Shawn

Been having all kinds of trouble installing cyassimp. I've tried many permutations and guesses, pyassimp, assimp, etc. None of them seem to work.

$ conda install -n my-fave-env -c menpo/label/master cyassimp
Fetching package metadata .............
Solving package specifications: 

PackageNotFoundError: Dependency missing in current osx-64 channels: 
  - cyassimp -> assimp 3.0.1270

Close matches found; did you mean one of these?

    assimp: cyassimp

Example program (basically straight from README.md)

import trimesh

mesh = trimesh.load_mesh([PLACE_STL_FILENAME_HERE])

# find groups of coplanar adjacent faces
facets, facets_area = mesh.facets(return_area=True)

# set each facet to a random color
for facet in facets:
  mesh.visual.face_colors[facet] = trimesh.visual.random_color()

# preview mesh in an opengl window if you installed pyglet with pip
mesh.show()

i tried to apply some boolean difference operation to a mesh. from a loaded model, i create a bounding cylinder and place a sphere within the scene. then i cut the sphere from the cylinder and rotate the cylinder.
after a couple a iterations, the mesh is surprisingly empty. and nearly all faces are away.

this test i did on ubuntu 16.10 with Python 2.7.12+ with blender and openscad

example.zip

Hello,

I got Qhull errors when I tried to generate the obb of the following meshes, do you have any suggestions?

Thanks in advance!

Vertices: [[19653.465921167503, 2901.982866568411, 22790.418736883457], [20611.665944905126, 2901.9828184502676, 22794.41878902774], [20612.93810257011, 2896.147085313453, 22489.677315260826], [19654.738078832495, 2896.1471334316, 22485.67726311654], [22372.26582808047, 6.382596408072594, 22857.21861265048], [22373.537985745457, 0.5468632712579691, 22552.47713888356], [21414.065752396516, 6.382729955398236, 22853.218558653392], [21415.337910061502, 0.5469968185836116, 22548.477084886475]]
Faces: [[2, 3, 1], [3, 0, 1], [1, 4, 5], [2, 1, 5], [7, 5, 6], [5, 4, 6], [6, 0, 7], [0, 3, 7], [3, 2, 7], [7, 2, 5], [6, 4, 1], [0, 6, 1]]

Vertices: [[18695.265921167505, 2901.982866568411, 22786.418736883457], [19653.46594490513, 2901.9828184502676, 22790.41878902774], [19654.738102570114, 2896.147085313453, 22485.677315260826], [18696.538078832495, 2896.1471334316, 22481.67726311654], [21414.065828080475, 6.382596408072594, 22853.21861265048], [21415.33798574546, 0.5468632712579691, 22548.47713888356], [20455.86575239652, 6.382729955398236, 22849.218558653396], [20457.137910061505, 0.5469968185836116, 22544.477084886475]]
Faces: [[2, 3, 1], [3, 0, 1], [1, 4, 5], [2, 1, 5], [7, 5, 6], [5, 4, 6], [6, 0, 7], [0, 3, 7], [3, 2, 7], [7, 2, 5], [6, 4, 1], [0, 6, 1]]

Vertices: [[17736.23126780341, 2901.4753593419673, 22780.023931916312], [18694.431251570077, 2901.475351664289, 22786.423959681106], [18696.466715963248, 2896.1036329803596, 22481.67809584849], [17738.266732196586, 2896.103640658043, 22475.278068083695], [20455.03128455919, 5.875370265509883, 22849.223785602746], [20457.066748952362, 0.503651581581277, 22544.477921770136]]
Faces: [[3, 1, 2], [0, 1, 3], [1, 4, 2], [4, 5, 2], [3, 4, 0], [3, 5, 4], [3, 2, 5], [0, 4, 1]]

Vertices: [[17736.50949556136, 2901.736576776612, 22780.02057682472], [20455.30942843662, 6.136314895439643, 22849.220251703195], [20457.0904373139, 0.525761342222264, 22544.477098053758], [17738.290504438646, 2896.1260232233985, 22475.27742317528], [19497.109334775803, 6.136461474350874, 22843.620245025544], [19498.890343653078, 0.5259079211334949, 22538.877091376104]]
Faces: [[0, 1, 3], [1, 2, 3], [1, 4, 5], [1, 5, 2], [3, 4, 0], [3, 5, 4], [5, 3, 2], [4, 1, 0]]

Hi,

I've not been able to load .ply files with vertex colours, neither ASCII nor binary. These files open fine in Meshlab, Blender and 3D Builder.

Basically what I am trying to do is load and view a mesh with vertex colours. Is that not something that is supported in trimesh or in the viewer? If I load an obj with vertex colours, it just displays the shape, no colours. Converting to .ply files is even less successful:

I'm getting various errors, depending on whether I use binary or ASCII ply files:

mesh = trimesh.load_mesh(r"mesh_ascii.ply")
Traceback (most recent call last):
  File "C:\Program Files (x86)\JetBrains\PyCharm 2017.1\helpers\pydev\_pydevd_bundle\pydevd_exec2.py", line 3, in Exec
    exec(exp, global_vars, local_vars)
  File "<input>", line 1, in <module>
  File "C:\Users\Patrik\Envs\eos-py\lib\site-packages\trimesh\constants.py", line 125, in timed
    result = method(*args, **kwargs)
  File "C:\Users\Patrik\Envs\eos-py\lib\site-packages\trimesh\io\load.py", line 111, in load_mesh
    file_type)
  File "C:\Users\Patrik\Envs\eos-py\lib\site-packages\trimesh\io\ply.py", line 35, in load_ply
    ply_ascii(elements, file_obj)
  File "C:\Users\Patrik\Envs\eos-py\lib\site-packages\trimesh\io\ply.py", line 225, in ply_ascii
    count].reshape((-1, rows)).astype(dtype)[:, 1:]
ValueError: invalid literal for int() with base 10: '0.188808'

Seems like on this one it expects integer colours, but they're given as float's?

Same mesh loading as binary ply:

Traceback (most recent call last):
  File "C:\Program Files (x86)\JetBrains\PyCharm 2017.1\helpers\pydev\_pydevd_bundle\pydevd_exec2.py", line 3, in Exec
    exec(exp, global_vars, local_vars)
  File "<input>", line 1, in <module>
  File "C:\Users\Patrik\Envs\eos-py\lib\site-packages\trimesh\constants.py", line 125, in timed
    result = method(*args, **kwargs)
  File "C:\Users\Patrik\Envs\eos-py\lib\site-packages\trimesh\io\load.py", line 111, in load_mesh
    file_type)
  File "C:\Users\Patrik\Envs\eos-py\lib\site-packages\trimesh\io\ply.py", line 37, in load_ply
    ply_binary(elements, file_obj)
  File "C:\Users\Patrik\Envs\eos-py\lib\site-packages\trimesh\io\ply.py", line 304, in ply_binary
    populate_listsize(file_obj, elements)
  File "C:\Users\Patrik\Envs\eos-py\lib\site-packages\trimesh\io\ply.py", line 267, in populate_listsize
    offset = np.dtype(prior_data).itemsize()
TypeError: 'int' object is not callable

I was getting various other errors as well with this mesh (after converting it from .obj to .ply with Meshlab) and then trying to load it with trimesh.

Are vertex colours not supported by trimesh?

I'm using Win 10, a virtualenv with python-3.6.1 (64-bit), numpy+mkl, scipy and Shapely from here, and pip install trimesh[easy].

Thanks

Thanks for this nice work.

I've used it in several models, but for one object, I got an error when I try to get the oriented bounding box:
Error in `python': free(): invalid pointer:

data:
'Faces': [[0, 1, 2],
[0, 2, 3],
[5, 2, 4],
[4, 2, 1],
[4, 6, 5],
[6, 7, 5],
[6, 0, 3],
[7, 6, 3],
[2, 5, 3],
[3, 5, 7],
[4, 1, 0],
[4, 0, 6]],
'Vertices': [[-1.8189894000000002e-12, 152.4, 0.0],
[-8.1854523e-11, 152.4, 2438.4],
[4572.0, 152.4, 2438.4],
[4572.0, 152.4, 0.0],
[185.32131, -152.4, 2438.4],
[4572.0, -152.4, 2438.4],
[185.32131, -152.4, 0.0],
[4572.0, -152.4, 0.0]],

In trimesh/ray/ray_triangle.py, function intersects_location.

rays parameter is not used anymore, but ray_origins and ray_directions

Weird example here.

I have a particular mesh that is watertight pre-rezero, and non-watertight post-rezero.

I'm not sure why this would occur! It seems that other meshes I try do not have this problem.

Here's the code:

import trimesh
import numpy as np

def main():
  m = trimesh.load_mesh("rezero-test.stl")
  watertight_rezero_test(m)

def watertight_rezero_test(m):
  print(m.is_watertight)

  m.rezero()

  print(m.is_watertight)


main()

The .stl file is here: https://drive.google.com/open?id=0B18jNduc8kjlQXROOVNCMlRENFU

Hey,

You're calling super with self.class in trimesh.scene.viewer.SceneViewer.flip() rather than SceneViewer. This means if I inherit from SceneViewer I'm forced to override flip to avoid the infinite recursion when it is called (as self.class is always going to be my inherited class).

Your package is generally awesome and you have called super correctly elsewhere within file so I'm sure its just an oversight.

Cheers,
Rob

Dear Mike,

Thank you for the prompt reply on my previous question. I have another question which might again be pyembree related. You commented briefly a potential issue in the ray_pyembree.py module:

# based on an internal tolerance of embree?
# 1e-4 definetly doesn't work
_ray_offset_distance = .01

The problem I am facing is following. By creating a mesh with a large coordinates values and offsetting a ray origin from the mesh surface by 0.001 or 0.01 the ray.intersection_any reports intersection (True) despite the ray origin is offset. This does not happen in all cases. When move the same mesh close to the Cartesian origin (0,0,0) there is no intersection; which is correct. I created a simple example which shows this misbehaviour. The mesh has a single rectangle composed of two triangles. Here is the code:

import numpy as np
import trimesh


def unit_normal_vector(p1, p2, p3):
    x, y, z = np.cross((p2 - p1), (p3 - p1))
    magnitude = np.sqrt(np.square(x) + np.square(y) + np.square(z))
    unit_normal_vector = list(np.array([x, y, z]) / magnitude)
    return unit_normal_vector

# Mesh first vertice 528980.85 183526.1 0
mesh = trimesh.load_mesh('./test_mesh.obj')

# Extract coordinates and determine a plane's unit normal vector
p1o, p2o, p3o = mesh.vertices[0:3]
unv = unit_normal_vector(p1o, p2o, p3o)

# Centroid of a rectangle's bottom edge elevated by 1m from the bottom.
cen = np.mean([p1o, p3o], axis=0)
cen[2] = 1

# Ray origin offset from the mesh rectangle by 0.01 along the unit normal vector
ep = cen + 0.01 * np.array(unv)

# Intersection returns True
intersection = mesh.ray.intersects_any([ep], [unv])
print(intersection)

# Mesh moved towards Cartesian origin (First vertice 80.85 26.1 0)
mesh_moved = trimesh.load_mesh('./test_mesh_moved.obj')

# Extract coordinates and determine a plane's unit normal vector
p1m, p2m, p3m = mesh_moved.vertices[0:3]
unv_moved = unit_normal_vector(p1m, p2m, p3m)

# Centroid of a rectangle's bottom edge elevated by 1m from the bottom.
cen_moved = np.mean([p1m, p3m], axis=0)
cen_moved[2] = 1

# Ray origin offset from the mesh rectangle by 0.001 along the unit normal vector
ep_moved = cen_moved + 0.001 * np.array(unv_moved)

# Intersection returns False (as it should be)
intersection_moved = mesh_moved.ray.intersects_any([ep_moved], [unv_moved])
print(intersection_moved)

test_mesh.obj

v 528980.85 183526.1 0
v 528980.85 183526.1 3
v 528982.3 183523.55 0
v 528982.3 183523.55 3
f 2 1 3
f 2 3 4

test_mesh_moved.obj

v 80.85 26.1 0
v 80.85 26.1 3
v 82.3 23.55 0
v 82.3 23.55 3
f 2 1 3
f 2 3 4

Best regards,
Ivan
p.s. Thank you very much for your effort in developing the trimesh library. It is a great library which I've been using a lot recently.

Hi Mike,

Thanks again for the excellent package :)

I was trying to extrude a 2D path today, so I did extruded = section2d.extrude(height=1), which returned 2 primitives. However, when I tried to export the primitives using extruded[0].export('test.stl'), an error shows up saying ValueError: extrude polygon not specified!. The complete code is below:

type(section2d)
Out[13]: trimesh.path.path.Path2D

extruded = section2d.extrude(height=1)

extruded[0].export('test.stl')
Traceback (most recent call last):

  File "<ipython-input-15-4ddd38bd4376>", line 1, in <module>
    extruded[0].export('test.stl')

  File "C:\Anaconda3\lib\site-packages\trimesh\base.py", line 1244, in export
    file_type = file_type)

  File "C:\Anaconda3\lib\site-packages\trimesh\io\export.py", line 31, in export_mesh
    log.info('Exporting %d faces as %s', len(mesh.faces), file_type.upper())

  File "C:\Anaconda3\lib\site-packages\trimesh\primitives.py", line 26, in faces
    self._create_mesh()

  File "C:\Anaconda3\lib\site-packages\trimesh\primitives.py", line 268, in _create_mesh
    mesh = creation.extrude_polygon(self.extrude_polygon,

  File "C:\Anaconda3\lib\site-packages\trimesh\primitives.py", line 244, in extrude_polygon
    raise ValueError('extrude polygon not specified!')

ValueError: extrude polygon not specified!

I do have meshpy installed. Any thoughts?

Thanks!

Shawn

PS: My end goal is to output a Path2D into an STL file with a single surface, so I was thinking about extrude with height=0. I'd really appreciate it if you think there is a less a hack-ish way as well!

Not an issue, just a question:

Is there any plan to implement CSG using blender instead of OpenSCAD? I've outsourced the CSG I've been doing to Blender, and it's been performing better for me.

I'm not sure if the speedup is just due to the opening time of OpenSCAD, or the actual speed of the CSG operation within it.

I am trying to perform boolean operations, however none of the interfaces seem to work properly. I have blender and openscad installed in my system but I get these errors:

SCAD:

> Traceback (most recent call last):
>   File "/home/tomas/Synology/Documents/Python/VPPy/VPPy-3/LPP.py", line 619, in <module>
>     d = geo.difference(sea, engine=engine)
>   File "/usr/local/lib/python2.7/dist-packages/trimesh/base.py", line 1029, in difference
>     engine = engine))
>   File "/usr/local/lib/python2.7/dist-packages/trimesh/boolean.py", line 19, in difference
>     result = _engines[engine](meshes, operation='difference')
>   File "/usr/local/lib/python2.7/dist-packages/trimesh/interfaces/scad.py", line 32, in boolean
>     return interface_scad(meshes, script)
>   File "/usr/local/lib/python2.7/dist-packages/trimesh/interfaces/scad.py", line 21, in interface_scad
>     result = scad.run('openscad $script -o $mesh_post')
>   File "/usr/local/lib/python2.7/dist-packages/trimesh/interfaces/generic.py", line 39, in run
>     check_call(command_run)
>   File "/usr/lib/python2.7/subprocess.py", line 506, in check_call
>     retcode = call(*popenargs, **kwargs)
>   File "/usr/lib/python2.7/subprocess.py", line 493, in call
>     return Popen(*popenargs, **kwargs).wait()
>   File "/usr/lib/python2.7/subprocess.py", line 679, in __init__
>     errread, errwrite)
>   File "/usr/lib/python2.7/subprocess.py", line 1249, in _execute_child
>     raise child_exception
> OSError: [Errno 2] No such file or directory

BLENDER:

> AL lib: pulseaudio.c:331: PulseAudio returned minreq > tlength/2; expect break up
> AL lib: pulseaudio.c:331: PulseAudio returned minreq > tlength/2; expect break up
> Traceback (most recent call last):
>   File "/tmp/tmpZC05uT", line 25, in <module>
>     use_mesh_modifiers = True)
>   File "/usr/lib/blender/scripts/modules/bpy/ops.py", line 180, in __call__
>     ret = op_call(self.idname_py(), None, kw)
> TypeError: Converting py args to operator properties: : keyword "use_mesh_modifiers" unrecognized
> 
> Blender quit

It would be useful to be able to find from a given point (specifying an optional search direction):

• Nearest vertex
• Nearest face
• Nearest face normal vector
• Nearest triangle center

There all seem easy enough to write up, but interpolation onto the surface would be maybe even more useful. Let me know what you think and I may try to work up a PR.