Module sverchok.utils.field.probe
Expand source code
# This file is part of project Sverchok. It's copyrighted by the contributors
# recorded in the version control history of the file, available from
# its original location https://github.com/nortikin/sverchok/commit/master
#
# SPDX-License-Identifier: GPL3
# License-Filename: LICENSE
import random
import numpy as np
from sverchok.utils.sv_logging import sv_logger
from sverchok.utils.kdtree import SvKdTree
BATCH_SIZE = 50
MAX_ITERATIONS = 1000
def _check_min_distance(v_new, vs_old, min_r):
if not vs_old:
return True
kdt = SvKdTree.new(SvKdTree.BLENDER, vs_old)
nearest, idx, dist = kdt.query(v_new)
if dist is None:
return True
ok = (dist >= min_r)
#if not ok:
# print(f"V {v_new} => {nearest}, {dist} >= {min_r}")
return ok
def _check_min_radius(point, old_points, old_radiuses, min_r):
if not old_points:
return True
old_points = np.array(old_points)
old_radiuses = np.array(old_radiuses)
point = np.array(point)
distances = np.linalg.norm(old_points - point, axis=1)
ok = (old_radiuses + min_r < distances).all()
return ok
def field_random_probe(field, bbox, count,
threshold=0, proportional=False, field_min=None, field_max=None,
min_r=0, min_r_field=None,
random_radius = False,
seed=0, predicate=None):
"""
Generate random points within bounding box, with distribution controlled (optionally) by a scalar field.
inputs:
* field: SvScalarField. Pass None to use uniform distribution.
* bbox: nested tuple: ((min_x, min_y, min_z), (max_x, max_y, max_z)).
* count: number of points to generate.
* threshold: do not generate points where value of scalar field is less than this value.
* proportional: if True, then density of points will be proportional to
values of the scalar field. Otherwise, values of the field will be used
only to not generate points in places where scalar field value is less than
threshold.
* field_min: (expected) minimum value of scalar field within the bounding box.
Mandatory if `proportional` is set to True.
* field_max: (expected) maximum value of scalar field within the bounding box.
Mandatory if `proportional` is set to True.
* min_r: minimum distance between generated points. Set to zero to disable this check.
* seed: random generator seed value.
* predicate: additional predicate to check if generated point is valid. Optional.
outputs:
list of vertices.
"""
if min_r != 0 and min_r_field is not None:
raise Exception("min_r and min_r_field can not be specified simultaneously")
if seed == 0:
seed = 12345
if seed is not None:
random.seed(seed)
b1, b2 = bbox
x_min, y_min, z_min = b1
x_max, y_max, z_max = b2
done = 0
generated_verts = []
generated_radiuses = []
iterations = 0
while done < count:
iterations += 1
if iterations > MAX_ITERATIONS:
sv_logger.error("Maximum number of iterations (%s) reached, stop.", MAX_ITERATIONS)
break
batch_xs = []
batch_ys = []
batch_zs = []
batch = []
left = count - done
max_size = min(BATCH_SIZE, left)
for i in range(max_size):
x = random.uniform(x_min, x_max)
y = random.uniform(y_min, y_max)
z = random.uniform(z_min, z_max)
batch_xs.append(x)
batch_ys.append(y)
batch_zs.append(z)
batch.append((x, y, z))
batch_xs = np.array(batch_xs)#[np.newaxis][np.newaxis]
batch_ys = np.array(batch_ys)#[np.newaxis][np.newaxis]
batch_zs = np.array(batch_zs)#[np.newaxis][np.newaxis]
batch = np.array(batch)
if field is None:
candidates = batch.tolist()
else:
values = field.evaluate_grid(batch_xs, batch_ys, batch_zs)
good_idxs = values >= threshold
if not proportional:
candidates = batch[good_idxs].tolist()
else:
candidates = []
for vert, value in zip(batch[good_idxs].tolist(), values[good_idxs].tolist()):
probe = random.uniform(field_min, field_max)
if probe <= value:
candidates.append(vert)
good_radiuses = []
if min_r == 0 and min_r_field is None:
good_verts = candidates
good_radiuses = [0 for i in range(len(good_verts))]
elif min_r_field is not None:
xs = np.array([p[0] for p in candidates])
ys = np.array([p[1] for p in candidates])
zs = np.array([p[2] for p in candidates])
min_rs = min_r_field.evaluate_grid(xs, ys, zs).tolist()
good_verts = []
for candidate, min_r in zip(candidates, min_rs):
if random_radius:
min_r = random.uniform(0, min_r)
if _check_min_radius(candidate, generated_verts + good_verts, generated_radiuses + good_radiuses, min_r):
good_verts.append(candidate)
good_radiuses.append(min_r)
else: # min_r != 0
good_verts = []
for candidate in candidates:
if _check_min_distance(candidate, generated_verts + good_verts, min_r):
good_verts.append(candidate)
good_radiuses = [1 for c in good_verts]
if predicate is not None:
pairs = [(vert, r) for vert, r in zip(good_verts, good_radiuses) if predicate(vert)]
good_verts = [p[0] for p in pairs]
good_radiuses = [p[1] for p in pairs]
generated_verts.extend(good_verts)
generated_radiuses.extend(good_radiuses)
done += len(good_verts)
return generated_verts, generated_radiusesFunctions
def field_random_probe(field, bbox, count, threshold=0, proportional=False, field_min=None, field_max=None, min_r=0, min_r_field=None, random_radius=False, seed=0, predicate=None)-
Generate random points within bounding box, with distribution controlled (optionally) by a scalar field.
inputs: * field: SvScalarField. Pass None to use uniform distribution. * bbox: nested tuple: ((min_x, min_y, min_z), (max_x, max_y, max_z)). * count: number of points to generate. * threshold: do not generate points where value of scalar field is less than this value. * proportional: if True, then density of points will be proportional to values of the scalar field. Otherwise, values of the field will be used only to not generate points in places where scalar field value is less than threshold. * field_min: (expected) minimum value of scalar field within the bounding box. Mandatory if
proportionalis set to True. * field_max: (expected) maximum value of scalar field within the bounding box. Mandatory ifproportionalis set to True. * min_r: minimum distance between generated points. Set to zero to disable this check. * seed: random generator seed value. * predicate: additional predicate to check if generated point is valid. Optional.outputs: list of vertices.
Expand source code
def field_random_probe(field, bbox, count, threshold=0, proportional=False, field_min=None, field_max=None, min_r=0, min_r_field=None, random_radius = False, seed=0, predicate=None): """ Generate random points within bounding box, with distribution controlled (optionally) by a scalar field. inputs: * field: SvScalarField. Pass None to use uniform distribution. * bbox: nested tuple: ((min_x, min_y, min_z), (max_x, max_y, max_z)). * count: number of points to generate. * threshold: do not generate points where value of scalar field is less than this value. * proportional: if True, then density of points will be proportional to values of the scalar field. Otherwise, values of the field will be used only to not generate points in places where scalar field value is less than threshold. * field_min: (expected) minimum value of scalar field within the bounding box. Mandatory if `proportional` is set to True. * field_max: (expected) maximum value of scalar field within the bounding box. Mandatory if `proportional` is set to True. * min_r: minimum distance between generated points. Set to zero to disable this check. * seed: random generator seed value. * predicate: additional predicate to check if generated point is valid. Optional. outputs: list of vertices. """ if min_r != 0 and min_r_field is not None: raise Exception("min_r and min_r_field can not be specified simultaneously") if seed == 0: seed = 12345 if seed is not None: random.seed(seed) b1, b2 = bbox x_min, y_min, z_min = b1 x_max, y_max, z_max = b2 done = 0 generated_verts = [] generated_radiuses = [] iterations = 0 while done < count: iterations += 1 if iterations > MAX_ITERATIONS: sv_logger.error("Maximum number of iterations (%s) reached, stop.", MAX_ITERATIONS) break batch_xs = [] batch_ys = [] batch_zs = [] batch = [] left = count - done max_size = min(BATCH_SIZE, left) for i in range(max_size): x = random.uniform(x_min, x_max) y = random.uniform(y_min, y_max) z = random.uniform(z_min, z_max) batch_xs.append(x) batch_ys.append(y) batch_zs.append(z) batch.append((x, y, z)) batch_xs = np.array(batch_xs)#[np.newaxis][np.newaxis] batch_ys = np.array(batch_ys)#[np.newaxis][np.newaxis] batch_zs = np.array(batch_zs)#[np.newaxis][np.newaxis] batch = np.array(batch) if field is None: candidates = batch.tolist() else: values = field.evaluate_grid(batch_xs, batch_ys, batch_zs) good_idxs = values >= threshold if not proportional: candidates = batch[good_idxs].tolist() else: candidates = [] for vert, value in zip(batch[good_idxs].tolist(), values[good_idxs].tolist()): probe = random.uniform(field_min, field_max) if probe <= value: candidates.append(vert) good_radiuses = [] if min_r == 0 and min_r_field is None: good_verts = candidates good_radiuses = [0 for i in range(len(good_verts))] elif min_r_field is not None: xs = np.array([p[0] for p in candidates]) ys = np.array([p[1] for p in candidates]) zs = np.array([p[2] for p in candidates]) min_rs = min_r_field.evaluate_grid(xs, ys, zs).tolist() good_verts = [] for candidate, min_r in zip(candidates, min_rs): if random_radius: min_r = random.uniform(0, min_r) if _check_min_radius(candidate, generated_verts + good_verts, generated_radiuses + good_radiuses, min_r): good_verts.append(candidate) good_radiuses.append(min_r) else: # min_r != 0 good_verts = [] for candidate in candidates: if _check_min_distance(candidate, generated_verts + good_verts, min_r): good_verts.append(candidate) good_radiuses = [1 for c in good_verts] if predicate is not None: pairs = [(vert, r) for vert, r in zip(good_verts, good_radiuses) if predicate(vert)] good_verts = [p[0] for p in pairs] good_radiuses = [p[1] for p in pairs] generated_verts.extend(good_verts) generated_radiuses.extend(good_radiuses) done += len(good_verts) return generated_verts, generated_radiuses