lavavu-osmesa 1.8.43__cp311-cp311-manylinux_2_24_x86_64.whl

lavavu/convert.py

@@ -0,0 +1,853 @@
+from __future__ import print_function
+"""
+Warning! EXPERIMENTAL:
+ these features and functions are under development, will have bugs,
+ and may be heavily modified in the future
+
+Tools for converting between 3D data types
+
+- Points to Volume
+- Triangles to OBJ file
+"""
+import numpy
+import os
+import sys
+
+def try_import(module_name, second_try=False):
+    """
+    Attempts to import a module, then runs pip install if not found and attempts again
+    """
+    import importlib
+    try:
+        m = importlib.import_module(module_name)
+        globals()[module_name] = m
+        return m
+    except (ImportError) as e:
+        if not second_try:
+            import subprocess
+            subprocess.check_call([sys.executable, "-m", "pip", "install", module_name])
+            return try_import(module_name, True)
+        else:
+            raise(e)
+
+def min_max_range(verts):
+    """
+    Get bounding box from a list of vertices
+    returns (min, max, range)
+    """
+    vmin = numpy.min(verts, axis=0)
+    vmax = numpy.max(verts, axis=0)
+    vrange = vmax - vmin
+    print("Bounding box ", (vmin, vmax), "Range ", vrange)
+    return (vmin, vmax, vrange)
+
+def default_sample_grid(vrange, res=8):
+    """Calculate sample grid fine enough to capture point details
+    resolution of smallest dimension will be minres elements
+    """
+    #If provided a full resolution, use that, otherwise will be interpreted as min res
+    if isinstance(res, (list,tuple)):
+        if len(list(res)) == 3:
+            return list(res)
+        res = res[0]
+    #Use bounding box range min
+    minr = numpy.min(vrange)
+    #res parameter is minimum resolution
+    factor = float(res) / float(minr)
+    RES = [int(factor*(vrange[0])), int(factor*(vrange[1])), int(factor*(vrange[2]))]   
+    print("Sample grid RES:",RES)
+    return RES
+
+def points_to_volume(verts, res=8, kdtree=False, normed=False, clamp=None, boundingbox=None):
+    """
+    Convert object vertices to a volume by interpolating points to a grid
+
+    - Vertex data only is used, so treated as a point cloud
+    - Can also be used to sample an irregular grid to a regular so volume render matches the grid dimensions
+    - Result is a density field that can be volume rendered
+
+    Default is to use numpy.histogramdd method, pass kdtree=True to use scipy.spatial.KDTree
+
+    TODO: support colour data too, converted density field becomes alpha channel
+
+    Returns
+    -------
+    values: numpy array of float32
+        The converted density field
+    boundingbox : 2,3
+        the minimum 3d vertex of the bounding box
+        the maximum 3d vertex of the bounding box
+
+    """
+    if kdtree:
+        return points_to_volume_tree(verts, res)
+    else:
+        return points_to_volume_histogram(verts, res, normed, clamp, boundingbox)
+
+def points_to_volume_histogram(verts, res=8, normed=False, clamp=None, boundingbox=None):
+    """
+    Using numpy.histogramdd to create 3d histogram volume grid
+    (Easily the fastest, but less control over output)
+    """
+    #Reshape to 3d vertices
+    verts = verts.reshape((-1,3))
+
+    #Get bounding box of swarm
+    vmin, vmax, vrange = min_max_range(verts)
+
+    #Minimum resolution to get ok sampling
+    if boundingbox is not None:
+        vmin = boundingbox[0]
+        vmax = boundingbox[1]
+        vrange = numpy.array(boundingbox[1]) - numpy.array(boundingbox[0])
+    RES = default_sample_grid(vrange, res)
+
+    #H, edges = numpy.histogramdd(verts, bins=RES)
+    if boundingbox is None:
+        H, edges = numpy.histogramdd(verts, bins=RES, normed=normed) #density=True for newer numpy
+    else:
+        rg = ((vmin[0], vmax[0]), (vmin[1], vmax[1]), (vmin[2], vmax[2])) #provide bounding box as range
+        H, edges = numpy.histogramdd(verts, bins=RES, range=rg, normed=normed) #density=True for newer numpy
+
+    #Reverse ordering X,Y,Z to Z,Y,X for volume data
+    values = H.transpose()
+
+    #Clamp [0,0.1] - optional (for a more binary output, points vs no points)
+    if clamp is not None:
+        values = numpy.clip(values, a_min=clamp[0], a_max=clamp[1])
+    
+    #Normalise [0,1] if using probability density
+    if normed:
+        values = values / numpy.max(values)
+    
+    return (values, vmin, vmax)
+
+def points_to_volume_tree(verts, res=8):
+    """
+    Using scipy.spatial.KDTree to find nearest points on grid
+
+    Much slower, but more control
+
+    TODO: control parameters
+    """
+    #Reshape to 3d vertices
+    verts = verts.reshape((-1,3))
+
+    #Get bounding box of swarm
+    lmin, lmax, lrange = min_max_range(verts)
+    print("Volume bounds: ",lmin.tolist(), lmax.tolist(), lrange.tolist())
+    
+    #Push out the edges a bit, will create a smoother boundary when we filter
+    #lmin -= 0.1*lrange
+    #lmax += 0.1*lrange
+    values = numpy.full(shape=(verts.shape[0]), dtype=numpy.float32, fill_value=1.0)
+
+    #Minimum resolution to get ok sampling
+    RES = default_sample_grid(lrange, res)
+
+    #Push out the edges a bit, will create a smoother boundary when we filter
+    cell = lrange / RES #Cell size
+    print("CELL:",cell)
+    lmin -= 2.0*cell
+    lmax += 2.0*cell
+
+    x = numpy.linspace(lmin[0], lmax[0] , RES[0])
+    y = numpy.linspace(lmin[1], lmax[1] , RES[1])
+    z = numpy.linspace(lmin[2], lmax[2] , RES[2])
+    print(x.shape, y.shape, z.shape)
+    Z, Y, X = numpy.meshgrid(z, y, x, indexing='ij') #ij=matrix indexing, xy=cartesian
+    XYZ = numpy.vstack((X,Y,Z)).reshape([3, -1]).T
+
+    print(lmin,lmax, XYZ.shape)
+
+    #KDtree for finding nearest neighbour points
+    from scipy import spatial
+    #sys.setrecursionlimit(1000)
+    #tree = spatial.KDTree(XYZ)
+    print("Building KDTree")
+    tree = spatial.cKDTree(XYZ)
+
+    #Outside distance to apply to grid points
+    MAXDIST = max(lrange) / max(RES) #Max cell size diagonal
+    print("Tree query, maxdist:",MAXDIST,max(lrange))
+
+    #Query all points, result is tuple: distances, indices
+    distances, indices = tree.query(verts, k=1) #Just get a single nearest neighbour for now
+
+    print("Calculate distances")
+    #Convert distances to [0,1] where 1=on grid and <= 0 is outside max range
+    distances = (MAXDIST - distances) / MAXDIST
+    print("Zero out of range distances")
+    distances *= (distances>0) #Zero negative elements by multiplication in-place
+
+    #Add the distances to the values at their nearest grid point indices
+    print("Compose distance field")
+    values = numpy.zeros(shape=(XYZ.shape[0]), dtype=numpy.float32)
+    values[indices] += distances
+    #Clip value max
+    print("Clip distance field")
+    values = values.clip(max=1.0)
+    
+    #Reshape to actual grid dims Z,Y,X... (not required but allows slicing)
+    XYZ = XYZ.reshape(RES[::-1] + [3])
+    values = values.reshape(RES[::-1])
+    print(XYZ.shape, values.shape)
+    
+    return (values, lmin, lmax)
+
+
+def points_to_volume_3D(vol, objects, res=8, kdtree=False, blur=0, pad=None, normed=False, clamp=None):
+    """
+    Interpolate points to grid and load into passed volume object
+
+    Given a list of objects and a volume object, convert a point cloud
+    from another object (or objects - list is supported) into a volume using
+    points_to_volume()
+    """
+    lv = vol.parent #Get the viewer from passed object
+    lv.hide(objects) #Hide the converted objects
+
+    #Get vertices from lavavu objects
+    pverts, bb_all = lv.get_all_vertices(objects)
+
+    #blur = False
+    #Use bounding box of full model?
+    #vmin, vmax, vrange = min_max_range([lv["min"], lv["max"]])
+    #bb_all == (vmin, vmax)
+    vdata, vmin, vmax = points_to_volume(pverts, res, kdtree, normed, clamp, bb_all)
+
+    if blur > 0:
+        if pad == None:
+            pad = int(blur*2)
+        if pad > 0:
+            print("Pad edges before blur", pad)
+            vdata = numpy.pad(vdata, pad, mode='constant')
+        print("Filter/blur distance field, sigma=%d" % blur)
+        from scipy.ndimage.filters import gaussian_filter
+        values = gaussian_filter(vdata, sigma=blur) #, mode='nearest')
+    else:
+        values = vdata #Need extra space at edges to use blur, so skip, or use pad() 
+    
+    print(numpy.min(values), numpy.max(values))
+    print(numpy.min(values), numpy.max(values))
+    
+    vol.values(values)
+    vol.vertices((vmin, vmax))
+
+def points_to_volume_4D(vol, objects, res=8, kdtree=False, blur=False, normed=False, clamp=None):
+    """
+    Interpolate points to grid at each timestep
+
+    Given a list of objects and a volume object, convert a time-varying point cloud
+    from another object (or objects - list is supported) into a volume using
+    points_to_volume_3D()
+    """
+    lv = vol.parent #Get the viewer from passed object
+
+    for step in lv.steps:
+        print("TIMESTEP:",step)
+        lv.timestep(step)
+        
+        points_to_volume_3D(vol, objects, res, kdtree, blur, normed, clamp)
+
+def colour2rgb(c):
+    return [c & 255, (c >> 8) & 255, (c >> 16) & 255]
+
+def colour2hex(rgb):
+    def padhex2(i):
+        s = hex(int(i))
+        return s[2:].zfill(2)
+    return "#" + padhex2(rgb[0]) + padhex2(rgb[1]) + padhex2(rgb[2])
+
+def _get_objects(source):
+    """
+    Returns a list of objects
+
+    If source is lavavu.Viewer() list contains all objects
+    If source is lavavu.Object() list contains that single object
+    """
+    if source.__class__.__name__ == 'Viewer':
+        return source.objects.list
+    elif not isinstance(source, list):
+        return [source]
+    else:
+        return source
+
+def export_OBJ(filepath, source, verbose=False, vertexcolours=True):
+    """
+    Export given object(s) to an OBJ file
+    Supports only triangle mesh object data
+
+    If source is lavavu.Viewer() exports all objects
+    If source is lavavu.Object() exports single object
+
+    Set vertexcolours to support writing R,G,B values with vertices
+    (not part of OBJ standard but supported in some software)
+    """
+    mtlfilename = os.path.splitext(filepath)[0] + '.mtl'
+    objects = _get_objects(source)
+    with open(filepath, 'w') as f, open(mtlfilename, 'w') as m:
+        f.write("# OBJ file\n")
+        offset = 1
+        for obj in objects:
+            if obj["visible"]:
+                f.write("g %s\n" % obj.name)
+                offset = _write_OBJ(f, m, filepath, obj, offset, verbose, vertexcolours)
+
+def _write_MTL(m, name, texture=None, diffuse=[1.0, 1.0, 1.0], ambient=None, specular=None, opacity=1.0, illum=None):
+    #http://paulbourke.net/dataformats/mtl/
+    #print("Writing MTL ", texture, diffuse, opacity, name)
+    mtl = "newmtl %s\n" % name
+    mtl += "Kd %.06f %.06ff %.06f\n" % (diffuse[0], diffuse[1], diffuse[2])
+    if ambient:
+        mtl += "Ka %.06f %.06ff %.06f\n" % (ambient[0], ambient[1], ambient[2])
+    if specular:
+        if illum is None:
+            illum = 2 #Highlight on
+        mtl += "Ks %.06f %.06ff %.06f\n" % (specular[0], specular[1], specular[2])
+        if len(specular) > 3:
+            mtl += "Ns %.06f\n" % specular[3]
+    mtl += "d %f\n" % (1.0 - opacity) #Dissolve: inverse of opacity
+    if illum is None:
+        illum = 1 #Default=1 = colour on, ambient on
+    mtl += "illum %d\n" % illum
+
+    if texture:
+        mtl += "map_Kd %s\n" % texture
+
+    mtl += '\n'
+    m.write(mtl)
+    mtl_line = "usemtl %s\n" % name
+    return mtl_line
+
+def _write_OBJ(f, m, filepath, obj, offset=1, verbose=False, vertexcolours=True):
+    mtl_line = ""
+    cmaptexcoords = []
+    colourdict = None
+    import re
+    name = re.sub(r"\s+", "", obj["name"])
+    colourcount = sum([len(c) for c in obj.data.colours])
+    #print(f"{colourcount=}")
+    if m and obj["texture"]:
+        fn = obj["texture"]
+        if fn == 'colourmap':
+            fn = 'texture.png'
+        #Write palette.png
+        obj.parent.palette('image', 'texture')
+        if verbose:
+            print("Writing texture mtl ", fn)
+        mtl_line = _write_MTL(m, name, texture=fn, opacity=obj["opacity"])
+
+    elif m and colourcount > 0 and not vertexcolours:
+        #Define material for each (slow for high colour/vertex count - could sort faces by material before writing?)
+        if colourcount < 10000:
+            colourdict = {}
+            #SLOW!
+
+            #Get unique https://stackoverflow.com/a/33197029/866759
+            allcolours = numpy.concatenate(obj.data.colours)
+            unique = numpy.unique(allcolours)
+
+            for c in unique:
+                rgb = colour2rgb(c)
+                cs = colour2hex(rgb)
+                colourdict[c] = (cs,rgb)
+            if verbose:
+                print("Writing mtl lib (colour list), unique colours:",len(unique))
+            for c in colourdict:
+                cs, rgb = colourdict[c]
+                mtl_line = _write_MTL(m, cs, diffuse=[rgb[0]/255.0, rgb[1]/255.0, rgb[2]/255.0], opacity=obj["opacity"])
+        else:
+            #Support this full RGB palette?
+            #TODO: generate texcoord data using formula below
+            #18bit RGB = 262144 colours
+            #Top-left = (0,0,0)
+            #Bottom-right = (255,255,255)
+            #(8x8 tiles of 64x64)
+            #G - slow, Z index = G
+            #R - midd, Y axis = R
+            #B - fast, X axis = B
+            #X = B//4
+            #Y = R//4
+            #Z = G//4
+            #zx = Z % 8
+            #zy = Z // 8
+            #x = zx * 64 + X
+            #y = zy * 64 + Y
+            #u = x / 512
+            #v = y / 512
+
+            #https://upload.wikimedia.org/wikipedia/commons/3/34/RGB_18bits_palette.png
+            palimg = "iVBORw0KGgoAAAANSUhEUgAAAgAAAAIACAIAAAB7GkOtAAAABGdBTUEAANjr9RwUqgAAACBjSFJNAACH"\
+                     "CgAAjAoAAPYWAACEzwAAczsAAOxVAAA6lwAAHU1girkoAAAGc0lEQVR42u3dQYoDMQxFwTZY93Duf8mQ"\
+                     "A8hg9yIIVTHbgU9vHsrG8/mZL/7iv/8+n2G//fbbb//V/vkA0JAAAAgAAAIAgAAAIAAACAAAAgCAAAAg"\
+                     "AAAIAAACAIAAACAAAAgAAAIAgAAAIAAACAAAAgDARQDCRwBwAQAgAAAIAAACAIAAACAAAAgAAAIAgAAA"\
+                     "IAAACAAAAgCAAAAgAAAIAAACAIAAACAAAAgAAAIAQB4Aj8IDuAAAEAAABAAAAQBAAAAQAAAEAAABAEAA"\
+                     "ABAAAAQAAAEAQAAAEAAABAAAAQBAAAAQAAAEAAABAGATAI/CA/QMwBrVC2a//fbbb//dfj8BAfS8AAQA"\
+                     "QAAAEAAABAAAAQBAAAAQAAAEAAABAEAAABAAAAQAAAEAQAAAEAAABAAAAQBAAAAQAADOAxA+AoALAAAB"\
+                     "AEAAABAAAAQAAAEAQAAAEAAABAAAAQBAAAAQAAAEAAABAEAAABAAAAQAAAEAQAAAEAAA8gB4FB7ABQCA"\
+                     "AAAgAAAIAAACAIAAACAAAAgAAAIAgAAAIAAACAAAAgCAAAAgAAAIAAACAIAAACAAAAgAAJsAeBQeoGcA"\
+                     "1qheMPvt77s/fH/7X+33ExBAzwtAAAAEAAABAEAAABAAAAQAAAEAQAAAEAAABAAAAQBAAAAQAAAEAAAB"\
+                     "AEAAABAAAAQAAAEA4DwA4SMAuAAAEAAABAAAAQBAAAAQAAAEAAABAEAAABAAAAQAAAEAQAAAEAAABAAA"\
+                     "AQBAAAAQAAAEAAABACAPgEfhAVwAAAgAAAIAgAAAIAAACAAAAgCAAAAgAAAIAAACAIAAACAAAAgAAAIA"\
+                     "gAAAIAAACAAAAgCAAACwCYBH4QF6BuAzqhfMfvvtt9/+u/1+AgLoeQEIAIAAACAAAAgAAAIAgAAAIAAA"\
+                     "CAAAAgCAAAAgAAAIAAACAIAAACAAAAgAAAIAgAAAIAAAnAcgfAQAFwAAAgCAAAAgAAAIAAACAIAAACAA"\
+                     "AAgAAAIAgAAAIAAACAAAAgCAAAAgAAAIAAACAIAAACAAAOQB8Cg8gAsAAAEAQAAAEAAABAAAAQBAAAAQ"\
+                     "AAAEAAABAEAAABAAAAQAAAEAQAAAEAAABAAAAQBAAAAQAAA2AfAoPEDPAKxRvWD222+//fbf7fcTEEDP"\
+                     "C0AAAAQAAAEAQAAAEAAABAAAAQBAAAAQAAAEAAABAEAAABAAAAQAAAEAQAAAEAAABAAAAQDgPADhIwC4"\
+                     "AAAQAAAEAAABAEAAABAAAAQAAAEAQAAAEAAABAAAAQBAAAAQAAAEAAABAEAAABAAAAQAAAEAIA+AR+EB"\
+                     "XAAACAAAAgCAAAAgAAAIAAACAIAAACAAAAgAAAIAgAAAIAAACAAAAgCAAAAgAAAIAAACAIAAALAJgEfh"\
+                     "AXoGYI3qBauten+n/fbbX3i/n4AAel4AAgAgAAAIAAACAIAAACAAAAgAAAIAgAAAIAAACAAAAgCAAAAg"\
+                     "AAAIAAACAIAAACAAAAgAAOcBCB8BwAUAgAAAIAAACAAAAgCAAAAgAAAIAAACAIAAACAAAAgAAAIAgAAA"\
+                     "IAAACAAAAgCAAAAgAAAIAAB5ADwKD+ACAEAAABAAAAQAAAEAQAAAEAAABAAAAQBAAAAQAAAEAAABAEAA"\
+                     "ABAAAAQAAAEAQAAAEAAABACATQA8Cg/QMwCfUb1g9ttvv/323+33ExBAzwtAAAAEAAABAEAAABAAAAQA"\
+                     "AAEAQAAAEAAABAAAAQBAAAAQAAAEAAABAEAAABAAAAQAAAEA4DwA4SMAuAAAEAAABAAAAQBAAAAQAAAE"\
+                     "AAABAEAAABAAAAQAAAEAQAAAEAAABAAAAQBAAAAQAAAEAAABACAPgEfhAVwAAAgAAAIAgAAAIAAACAAA"\
+                     "AgCAAAAgAAAIAAACAIAAACAAAAgAAAIAgAAAIAAACAAAAgCAAACwCYBH4QF6BmCN6gWz33777f+PKP/9"\
+                     "/QQE0PMCEAAAAQBAAAAQAAAEAAABAEAAABAAAAQAAAEAQAAAEAAABAAAAQBAAAAQAAAEAAABAEAAADgP"\
+                     "QPgIAC4AAAQAAAEAQAAAEAAABAAAAQBAAAAQAAAEAAABAEAAABAAAAQAAAEAQAAAEAAABAAAAQBAAADI"\
+                     "A+BReAAXAAACAIAAACAAAAgAAAIAgAAAIAAACAAAAgCAAAAgAAAIAAACAIAAACAAAAgAAAIAgAAAIAAA"\
+                     "bALgUXiAlr7Y4BnEOVAUKwAAAABJRU5ErkJggg=="
+            texfn = os.path.splitext(filepath)[0] + '.png'
+            import base64
+            with open(texfn, "wb") as fp:
+                fp.write(base64.b64decode(palimg))
+            mtl_line = _write_MTL(m, "palette_rgb", texture=texfn, opacity=obj["opacity"])
+    elif m and "colour" in obj:
+        #print("Writing mtl lib (default colour)")
+        c = obj.parent.parse_colour(obj["colour"])
+        mtl_line = _write_MTL(m, 'default-' + name, diffuse=c, opacity=obj["opacity"])
+
+    for o,data in enumerate(obj):
+        if verbose: print("[%s] element %d of %d" % (obj.name, o+1, len(obj.data.vertices)))
+        verts = data.vertices.reshape((-1,3))
+        if len(verts) == 0:
+            if verbose: print("No vertices")
+            continue
+        f.write("o Surface_%d\n" % o)
+        #f.write("o %s\n" % obj.name)
+        if m: f.write("mtllib " + os.path.splitext(os.path.basename(filepath))[0] + ".mtl\n")
+        indices = data.indices.reshape((-1,3))
+        normals = data.normals.reshape((-1,3))
+        texcoords = data.texcoords.reshape((-1,2))
+        #Calculate texcoords from colour values?
+        if len(texcoords) == 0 and obj["texture"] == 'colourmap':
+            label = obj["colourby"]
+            if isinstance(label,int):
+                #Use the given label index
+                sets = list(datasets.keys())
+                label = sets[label]
+            elif len(label) == 0:
+                #Use the default label
+                label = 'values'
+            valdata = data[label]
+            if len(valdata) >= o+1:
+                #Found matching value array
+                v = valdata
+                #Normalise [0,1]
+                texcoords = (v - numpy.min(v)) / numpy.ptp(v)
+                #Add 2nd dimension (not actually necessary,
+                #tex coords can by 1d but breaks some loaders (meshlab)
+                zeros = numpy.zeros((len(texcoords)))
+                texcoords = numpy.vstack((texcoords,zeros)).reshape([2, -1]).transpose()
+
+        #Colours?
+        cv0 = ""
+        vperc = 1
+        if len(data.colours):
+            vperc = int(verts.shape[0] / len(data.colours))
+
+        if verbose: print("- Writing vertices:",verts.shape)
+        for vi,v in enumerate(verts):
+            #Vertex colour with vertex? (only if flag passed)
+            if vertexcolours and len(data.colours):
+                #if vi%10==0: print(len(data.colours), len(verts), vperc, vi)
+                c = data.colours[len(data.colours) // vperc - 1]
+                rgb = colour2rgb(c)
+                f.write("v %.6f %.6f %.6f %.6f %.6f %.6f\n" % (v[0], v[1], v[2], rgb[0]/255.0, rgb[1]/255.0, rgb[2]/255.0))
+            else:
+                f.write("v %.6f %.6f %.6f\n" % (v[0], v[1], v[2]))
+        if verbose: print("- Writing normals:",normals.shape)
+        for n in normals:
+            f.write("vn %.6f %.6f %.6f\n" % (n[0], n[1], n[2]))
+        if verbose: print("- Writing texcoords:",texcoords.shape)
+        if len(texcoords.shape) == 2:
+            for t in texcoords:
+                f.write("vt %.6f %.6f\n" % (t[0], t[1]))
+        else:
+            for t in texcoords:
+                f.write("vt %.6f\n" % t)
+
+        #Face elements v/t/n v/t v//n
+        f.write(mtl_line)
+        if len(normals) and len(texcoords):
+            if verbose: print("- Writing faces (v/t/n):",indices.shape)
+        elif len(texcoords):
+            if verbose: print("- Writing faces (v/t):",indices.shape)
+        elif len(normals):
+            if verbose: print("- Writing faces (v//n):",indices.shape)
+        else:
+            if verbose: print("- Writing faces (v):",indices.shape)
+        if verbose: print("- Colours :",data.colours.shape)
+        if verbose: print("- Indices :",indices.shape)
+
+        for n,i in enumerate(indices):
+            if verbose and n%1000==0: print(".", end=''); sys.stdout.flush()
+            i0 = i[0]+offset
+            i1 = i[1]+offset
+            i2 = i[2]+offset
+            #Use mtl colours?
+            if colourdict:
+                ci = int(i[0] / vperc)
+                cv1 = data.colours[ci]
+                if cv0 != cv1 and cv1 in colourdict:
+                    f.write("usemtl " + colourdict[cv1][0] + "\n")
+                    cv0 = cv1
+
+            if len(normals) and len(texcoords):
+                f.write("f %d/%d/%d %d/%d/%d %d/%d/%d\n" % (i0, i0, i0, i1, i1, i1, i2, i2, i2))
+            elif len(texcoords):
+                f.write("f %d/%d %d/%d %d/%d\n" % (i0, i0, i1, i1, i2, i2))
+            elif len(normals):
+                f.write("f %d//%d %d//%d %d//%d\n" % (i0, i0, i1, i1, i2, i2))
+            else:
+                f.write("f %d %d %d\n" % (i0, i1, i2))
+        if verbose: print()
+
+        offset += verts.shape[0]
+    return offset
+
+def export_PLY(filepath, source, binary=True):
+    """
+    Export given object(s) to a PLY file
+    Supports points or triangle mesh object data
+
+    If source is lavavu.Viewer() exports all objects
+    If source is lavavu.Object() exports single object
+
+    Parameters
+    ----------
+    filepath : str
+        Output file to write
+    source : lavavu.Viewer or lavavu.Object
+        Where to get object data to export
+    binary : boolean
+        Write vertex/face data as binary, default True
+    """
+    objects = _get_objects(source)
+    with open(filepath, mode='wb') as f:
+        voffset = 0
+        foffset = 0
+        #First count vertices, faces
+        fc = 0
+        vc = 0
+        for obj in objects:
+            for o,data in enumerate(obj):
+                vc += len(data.vertices) #Vertices now in shape (N,3) so len returns correct count
+                fc += len(data.indices) // 3
+
+        #Pass the counts first
+        vertex = None
+        face = None
+        print(vc, " vertices, ", fc, " faces")
+        for obj in objects:
+            for o,data in enumerate(obj):
+                print("[%s] element %d of %d, type %s" % (obj.name, o+1, len(obj.data), data.type))
+                #print("OFFSETS:",voffset,foffset)
+                verts = data.vertices.reshape((-1,3))
+                if len(verts) == 0:
+                    print("No vertices")
+                    return
+                if len(verts) != vc:
+                    print("Vertex count error!", len(verts), vc, verts.shape)
+                    return
+                indices = data.indices.reshape((-1,3))
+                normals = data.normals.reshape((-1,3))
+                texcoords = data.texcoords.reshape((-1,2))
+
+                vperc = 0
+                cperf = 0
+                if len(data.colours):
+                    vperc = int(verts.shape[0] / len(data.colours))
+                    C = data.colours
+                    #print("COLOURS:",len(C),C.shape, verts.shape[0], verts.shape[0] / len(data.colours), vperc)
+
+                if data.type != 'points':
+                    #Per face colours, or less
+                    if face is None:
+                        if vperc and vperc < len(verts):
+                            #cperf = int(indices.shape[0] / len(data.colours))
+                            face = numpy.zeros(shape=(fc), dtype=[('vertex_indices', 'i4', (3,)), ('red', 'u1'), ('green', 'u1'), ('blue', 'u1')])
+                        else:
+                            face = numpy.zeros(shape=(fc), dtype=[('vertex_indices', 'i4', (3,))])
+                        print("FACE:",face.dtype)
+
+                    for i,idx in enumerate(indices):
+                        if i%1000==0: print(".", end=''); sys.stdout.flush()
+                        if vperc and vperc < len(verts):
+                            #Have colour, but less than vertices, apply to faces
+                            ci = idx[0] // vperc
+                            c = data.colours[ci]
+                            rgb = colour2rgb(c)
+                            face[i+foffset] = ([idx[0]+voffset, idx[1]+voffset, idx[2]+voffset], rgb[0], rgb[1], rgb[2])
+                        else:
+                            face[i+foffset] = ([idx[0]+voffset, idx[1]+voffset, idx[2]+voffset])
+                    print()
+
+                #Construct and write vertex elements
+                if vertex is None:
+                    #Setup vertex array based on first object element
+                    D = [('x', 'f4'), ('y', 'f4'), ('z', 'f4')]
+                    if normals.shape[0] == verts.shape[0]:
+                        D += [('nx', 'f4'), ('ny', 'f4'), ('nz', 'f4')]
+                    if texcoords.shape[0] == verts.shape[0]:
+                        D += [('s', 'f4'), ('t', 'f4')]
+                    if vperc and vperc == 1:
+                        D += [('red', 'u1'), ('green', 'u1'), ('blue', 'u1')]
+                    print("VERTEX:",D)
+                    vertex = numpy.zeros(shape=(vc), dtype=D)
+
+                for i,v in enumerate(verts):
+                    #if i%1000==0:
+                    #    print("vert index",i,vperc)
+                    if i%1000==0: print(".", end=''); sys.stdout.flush()
+                    E = [v[0], v[1], v[2]]
+                    if normals.shape[0] == verts.shape[0]:
+                        N = normals[i]
+                        E += [N[0], N[1], N[2]]
+                    if texcoords.shape[0] == verts.shape[0]:
+                        T = texcoords[i]
+                        E += [T[0], T[1]]
+                    if vperc and vperc == 1:
+                        c = data.colours[i]
+                        rgb = colour2rgb(c)
+                        E += [rgb[0], rgb[1], rgb[2]]
+                    vertex[i+voffset] = tuple(E)
+                print()
+
+                #Update offsets : number of vertices / faces added
+                voffset += verts.shape[0]
+                foffset += indices.shape[0]
+
+        import plyfile
+        #vertex = numpy.array(vertex, dtype=vertex.dtype)
+        els = []
+        els.append(plyfile.PlyElement.describe(vertex, 'vertex'))
+        if face is not None:
+            els.append(plyfile.PlyElement.describe(face, 'face'))
+
+        #Write, text or binary
+        if binary:
+            print("Writing binary PLY data")
+            plyfile.PlyData(els).write(f)
+        else:
+            print("Writing ascii PLY data")
+            plyfile.PlyData(els, text=True).write(f)
+
+def _get_PLY_colours(element):
+    """
+    Extract colour data from PLY element and return as a numpy rgba array
+    """
+    r = None
+    g = None
+    b = None
+    a = None
+    #print(element.properties)
+    for prop in element.properties:
+        #print(prop,prop.name,prop.dtype)
+        if 'red' in prop.name: r = element[prop.name]
+        if 'green' in prop.name: g = element[prop.name]
+        if 'blue' in prop.name: b = element[prop.name]
+        if 'alpha' in prop.name: a = element[prop.name]
+
+    if r is not None and g is not None and b is not None:
+        if a is None:
+            a = numpy.full(r.shape, 255)
+        #return numpy.array([r, g, b, a])
+        return numpy.vstack((r,g,b,a)).reshape([4, -1]).transpose()
+
+    return None
+
+def plot_PLY(lv, filename):
+    """
+    Plot triangles from a PlyData instance. Assumptions:
+    has a 'vertex' element with 'x', 'y', and 'z' properties,
+    has a 'face' element with an integral list property 'vertex_indices',
+    all of whose elements have length 3.
+    """
+    import plyfile
+    plydata = plyfile.PlyData.read(filename)
+
+    x = plydata['vertex']['x']
+    y = plydata['vertex']['y']
+    z = plydata['vertex']['z']
+    V = numpy.vstack((x,y,z)).reshape([3, -1]).transpose()
+    #V = numpy.array([x, y, z])
+    #print("VERTS:", V.shape)
+
+    vp = []
+    for prop in plydata['vertex'].properties:
+        vp.append(prop.name)
+        print(prop.name)
+    
+    N = None
+    if 'nx' in vp and 'ny' in vp and 'nz' in vp:
+        nx = plydata['vertex']['nx']
+        ny = plydata['vertex']['ny']
+        nz = plydata['vertex']['nz']
+        N = numpy.vstack((nx,ny,nz)).reshape([3, -1]).transpose()
+    
+    T = None
+    if 's' in vp and 't' in vp:
+        s = plydata['vertex']['s']
+        t = plydata['vertex']['t']
+        T = numpy.vstack((s,t)).reshape([2, -1]).transpose()
+
+    C = _get_PLY_colours(plydata['vertex'])
+
+    if 'face' in plydata:
+        #print(plydata['face'])
+        #Face colours?
+        if C is None:
+            C = _get_PLY_colours(plydata['face'])
+
+        tri_idx = plydata['face']['vertex_indices']
+        idx_dtype = tri_idx[0].dtype
+        tri_idx = numpy.asarray(tri_idx).flatten()
+        #print(type(tri_idx),idx_dtype,tri_idx.shape)
+        #print(tri_idx)
+
+        triangles = numpy.array([t.tolist() for t in tri_idx]).flatten()
+        #triangles = numpy.fromiter(tri_idx, [('data', idx_dtype, (3,))], count=len(tri_idx))['data']
+
+        return lv.triangles(vertices=V, indices=triangles, colours=C, normals=N, texcoords=T)
+    else:
+        return lv.points(vertices=V, colours=C, normals=N, texcoords=T)
+
+def export_any(filepath, source, name=None):
+    """
+    Export given object(s) to a file format supproted by trimesh, eg: GLTF or GLB file
+    See: https://trimsh.org/trimesh.exchange.html
+    Requires "trimesh" module
+    Supports triangle mesh object data
+
+    If source is lavavu.Viewer() exports all objects
+    If source is lavavu.Object() exports single object
+
+    Parameters
+    ----------
+    filepath : str
+        Output file to write
+    source : lavavu.Viewer or lavavu.Object
+        Where to get object data to export
+    """
+    #TODO: support points, lines
+    try_import('trimesh')
+
+    objects = _get_objects(source)
+    scene = trimesh.Scene()
+    facect = 0
+    for obj in objects:
+        #Use object name to export unless name provided
+        oname = name
+        if oname is None:
+            oname = obj["name"]
+        for i,e in enumerate(obj.data):
+            #print(e)
+            meshdict = {}
+            meshdict["vertices"] = e.vertices.reshape(-1,3)
+            meshdict["vertex_normals"] = e.normals.reshape(-1,3)
+            meshdict["faces"] = e.indices.reshape(-1,3)
+            facect += e.indices.shape[0]
+            if not len(meshdict["faces"]):
+                print("Empty",e)
+                continue
+            colours = e.colours
+            #print("ColourCount",len(colours))
+            if len(colours) == len(meshdict["vertices"]):
+                #View int32 into uint8 bytes
+                meshdict["vertex_colors"] = colours.view('uint8').reshape((-1,4))
+
+            mesh = trimesh.load_mesh(meshdict)
+            scene.add_geometry(mesh, geom_name=oname + '#' + str(i))
+
+            if len(colours) <= 1:
+                #If we don't set a default material colour, trimesh import will set default colour for every vertex
+                colour = obj.parent.parse_colour(obj["colour"])
+                colour = numpy.array(colour*255, dtype=numpy.uint8)
+                if len(colours) > 0:
+                    colour = colours.view('uint8').reshape((-1,4))[0]
+                if obj["opacity"] < 1.0:
+                    colour[3] *= obj["opacity"]
+                #print("Single colour: export as material", colour)
+                mesh.visual = trimesh.visual.TextureVisuals(material=trimesh.visual.material.PBRMaterial())
+                mesh.visual.material.baseColorFactor = colour
+
+    if facect > 0:
+        scene.export(file_obj=filepath)
+    else:
+        print("No triangle facets to export")
+
+    """
+    #Try GLTF ascii?
+    gltf = trimesh.exchange.gltf.export_gltf(scene)
+    import json
+    print("Mesh loaded, exporting GLTF")
+    gltf = trimesh.exchange.gltf.export_gltf(mesh) #, merge_buffers=True)
+    #print(gltf.keys())
+    #print(gltf['model.gltf'])
+    for fn in gltf:
+        mode = 'w'
+        #if ".bin" in fn: mode = 'wb'
+        with open(fn, 'wb') as f:
+            print("writing gltf component:" + fn)
+            f.write(gltf[fn])
+    """
+
+def read_any(filepath, lv):
+    """
+    Load using trimesh, supports GLTF etc
+
+    See: https://trimsh.org/trimesh.exchange.html
+    Requires "trimesh" module
+    """
+    #TODO: support points, lines
+    try_import('trimesh')
+    scene = trimesh.load(filepath)
+    tris = None
+    for name in scene.geometry.keys():
+        geometry = scene.geometry[name]
+        idx = 0
+        if '#' in name:
+            name,idx = name.split('#')
+            idx = int(idx)
+
+        if idx == 0:
+            tris = lv.triangles(name)
+        else:
+            tris.append()
+
+        tris.vertices(geometry.vertices)
+        tris.normals(geometry.vertex_normals)
+        tris.indices(geometry.faces)
+
+        #adjacency_matrix = geometry.edges_sparse
+        #print(geometry)
+        #print(geometry.visual)
+
+        #Load vertex colours if available
+        if hasattr(geometry.visual, "vertex_colors"):
+            #print("HAVE VERTEX COLOURS",len(geometry.visual.vertex_colors))
+            tris.colours(geometry.visual.vertex_colors)
+        #Load single colour material
+        elif hasattr(geometry.visual, "material"):
+            #print("HAVE MATERIAL")
+            #print(geometry.visual.material)
+            #print(geometry.visual.material.baseColorFactor)
+            if idx == 0:
+                #Can set as prop, but only works for single el
+                tris["colour"] = geometry.visual.material.baseColorFactor
+            else:
+                tris.colours(geometry.visual.material.baseColorFactor)
+
+    return tris