def classify(centroids: Iterable[AnyVec]):
new_clusters: Dict[AnyVec, List[AnyVec]] = defaultdict(list)
tree = RTree(centroids)
for point in points:
nn, _ = tree.nearest_neighbor(point)
new_clusters[nn].append(point)
return new_clusters
def test_Vec2_compatibility():
tree = RTree([Vec2(x, 0) for x in range(100)], max_node_size=5)
bbox = BoundingBox([(45, 0, 0), (55, 0, 0)])
points = list(tree.points_in_bbox(bbox))
assert len(points) == 11
expected_x_coords = set(range(45, 56))
x_coords = set(int(p.x) for p in points)
assert x_coords == expected_x_coords
assert any(isinstance(p, Vec2) for p in points)
def dbscan(
points: List[AnyVec],
*,
radius: float,
min_points: int = 4,
rtree: RTree = None,
max_node_size: int = 5,
) -> List[List[AnyVec]]:
"""DBSCAN clustering.
https://en.wikipedia.org/wiki/DBSCAN
Args:
points: list of points to cluster
radius: radius of the dense regions
min_points: minimum number of points that needs to be within the
`radius` for a point to be a core point (must be >= 2)
rtree: optional RTree
max_node_size: max node size for internally created RTree
Returns:
list of clusters, each cluster is a list of points
.. versionadded:: 0.18
"""
if min_points < 2:
raise ValueError("min_points must be >= 2")
if rtree is None:
rtree = RTree(points, max_node_size)
clusters: List[Set[AnyVec]] = []
point_set = set(points)
while len(point_set):
point = point_set.pop()
todo = {point}
cluster = {point} # the cluster has only a single entry if noise
clusters.append(cluster)
while len(todo):
chk_point = todo.pop()
neighbors = set(rtree.points_in_sphere(chk_point, radius))
if len(neighbors) < min_points:
continue
cluster.add(chk_point)
point_set.discard(chk_point)
todo |= neighbors.intersection(point_set)
return [list(cluster) for cluster in clusters]
def profile_rtree_contains_all_points(repeat: int, max_size: int):
log = []
for size in range(100, 2000, 100):
points = random_points(size, 50.0)
tree = RTree(points, max_size)
name = f"RTree({size}, {max_size})"
t0 = profile(
profile_tree_contains_points, repeat, tree, points
)
time_str = f"{t0:6.2f}s"
print(f"{name} contains all points, {repeat}x , {time_str}")
log.append((size, t0))
return log
def profile_rtree_nearest_neighbor(repeat: int, max_size: int):
log = []
for size in range(100, 2000, 100):
points = random_points(size, 50.0)
tree = RTree(points, max_size)
name = f"RTree({size}, {max_size})"
search_points = random_points(100, 50.0)
t0 = profile(
profile_tree_nearest_neighbor, repeat, tree, search_points
)
time_str = f"{t0:6.2f}s"
print(f"{name} nearest neighbor, {repeat}x , {time_str}")
log.append((size, t0))
return log
def test_from_two_points(self):
tree = RTree([Vec3(1, 2, 3), Vec3(3, 2, 1)])
assert len(tree) == 2
def test_iter_tree(self):
point = Vec3(1, 2, 3)
tree = RTree([point])
assert list(tree) == [point]
def test_contains_point(self):
point = Vec3(1, 2, 3)
tree = RTree([point])
assert tree.contains(point)
def test_avg_methods_return_0_for_too_small_trees():
tree = RTree([Vec3()])
assert tree.avg_leaf_size() == 0.0
assert tree.avg_spherical_envelope_radius() == 0.0
assert tree.avg_nn_distance() == 0.0
def test_avg_nn_distance_of_random_points():
tree = RTree([Vec3.random(100) for _ in range(100)])
nn_dist = tree.avg_nn_distance()
assert nn_dist > 10.0
def test_contains_all_random_points(self):
points = [Vec3.random(50) for _ in range(100)]
tree = RTree(points, 5)
for point in points:
assert tree.contains(point) is True
def test_average_leaf_size_of_random_points():
tree = RTree([Vec3.random(100) for _ in range(100)])
size = tree.avg_leaf_size()
assert size > 10.0
def test_collect_leafs():
tree = RTree([Vec3.random(100) for _ in range(100)])
assert sum(len(leaf) for leaf in rtree.collect_leafs(tree._root)) == 100
def profile_build_time_random_rtree(count: int, points, max_node_size: int):
for _ in range(count):
RTree(points, max_node_size)
def test_store_duplicate_points(self):
p = Vec3(1, 2, 3)
tree = RTree([p, p])
assert len(tree) == 2
def tree(self):
return RTree([Vec3(x, 0, 0) for x in range(100)], max_node_size=5)
def test_avg_spherical_envelope_radius_of_random_points():
tree = RTree([Vec3.random(100) for _ in range(100)])
radius = tree.avg_spherical_envelope_radius()
assert radius > 10.0
def test_can_not_build_empty_tree():
with pytest.raises(ValueError):
RTree([])
def test_from_one_point(self):
tree = RTree([Vec3(1, 2, 3)])
assert len(tree) == 1
