Blender k-means fractions

README.md

K-means Fractions Blender Script

This script splits a mesh in pow(CLUSTER_COUNT, ITERATION_COUNT) fractions in pseudo-random shapes.

1. Select an object
2. Distribute an amount of random points (SEED_POINT_COUNT) in the mesh of that object
3. K-means the points into k clusters (CLUSTER_COUNT)
4. Find the convex hull of the clusters, creating meshes
5. Remove original object
6. Go to step 1

Inspired by Matt DesLauriers' tweet: https://twitter.com/mattdesl/status/945728639496310786

Installing dependencies

The commends below work if you have pip (a Python package manager) installed. If not, see installing pip for Blender's Python.

• python -m pip install sklearn

How to use

• Select any object in the 3D View
• Run the script

Possible improvements

• Instead of picking the first selected object, iterate through all selected objects

create-kmeans-fractions.py

import bpy
import itertools
import mathutils

from mathutils import Vector
from sklearn.cluster import KMeans

CLUSTER_COUNT = 4
SEED_POINT_COUNT = 1024 * 4
ITERATION_COUNT = 3

def lerp(a, b, t):
	return t * (b - a) + a


def is_point_in_mesh(point, object):
	axes = [Vector((1, 0, 0)), Vector((0, 1, 0)), Vector((0, 0, 1))]

	is_outside = False

	for axis in axes:
		count = 0
		orig = point

		while True:
			result, location, normal, index = object.ray_cast(orig, orig + axis * 10000.0)

			if index == -1: break

			count += 1
			orig = location + axis * 0.00001
		if count % 2 == 0:
			is_outside = True
			break

	return not is_outside


def get_bounding_box_component(bounding_box, component_index):
	return list(map(lambda v: v[component_index], bounding_box))


def unselect_all():
	for obj in bpy.context.selected_objects:
			obj.select = False


def get_bounding_box_min_max(bounding_box):
	x = get_bounding_box_component(bounding_box, 0)
	y = get_bounding_box_component(bounding_box, 1)
	z = get_bounding_box_component(bounding_box, 2)

	vmin = Vector((min(x), min(y), min(z)))
	vmax = Vector((max(x), max(y), max(z)))
	return (vmin, vmax)


def generate_points(point_count, object):
	min, max = get_bounding_box_min_max(object.bound_box)
	points = []

	for x in range(0, point_count):
		points.append(Vector((
			lerp(min.x, max.x, mathutils.noise.random()),
			lerp(min.y, max.y, mathutils.noise.random()),
			lerp(min.z, max.z, mathutils.noise.random())
		)))

	return list(filter(lambda v: is_point_in_mesh(v, object), points))


def get_clusters(points, n_clusters=2):
	clusters = list(map(lambda c: [], range(0, n_clusters)))
	cluster_indexes = KMeans(n_clusters=n_clusters).fit(points).predict(points)

	for i in range(0, len(points)):
		point = points[i]
		cluster_index = cluster_indexes[i]
		clusters[cluster_index].append(point)

	return clusters


def replace_object(object):
	objects = []
	points = generate_points(SEED_POINT_COUNT, object)
	clusters = get_clusters(points, CLUSTER_COUNT)

	for cluster in clusters:
		unselect_all()

		mesh = bpy.data.meshes.new('fractal')
		mesh.from_pydata(cluster, [], [])

		fractal = bpy.data.objects.new('fractal', mesh)
		fractal.location = bpy.context.scene.cursor_location
		fractal.active_material = object.active_material
		
		bpy.context.scene.objects.link(fractal)
		bpy.context.scene.update()

		bpy.context.scene.objects.active = fractal
		fractal.select = True

		bpy.ops.object.mode_set(mode='EDIT')
		bpy.ops.mesh.convex_hull()
		bpy.ops.object.mode_set(mode='OBJECT')
		
		objects.append(fractal)
	
	bpy.context.scene.objects.unlink(object)
	
	return objects


def main():
	if len(bpy.context.selected_objects) == 0:
		console.error('No object selected')
		return
	
	objects = [bpy.context.selected_objects[0]]

	for i in range(0, ITERATION_COUNT):
		newObjects = list(map(lambda o: replace_object(o), objects))
		objects = list(itertools.chain(*newObjects))


main()

installing-pip.md

Installing pip for Blender's Python

• Download get-pip.py
• Find the python in your Blender installation (Blender Foudation/Blender/2.x/python/bin)
• Open a terminal and run python get-pip.py

Note: in case you run into errors, you might have to change the directory access settings to Blender\2.79\python\lib\site-packages, or run above script as administrator.

Now you can run pip via path/to/Blender/2.x/python/bin/python -m pip.