def test_impl():
n = 20
arr = np.ones((n, n))
val = 0
for idx in pndindex(arr.shape):
val += idx[0] * idx[1]
return val
def do_scale(X, maxv, nthr):
with numba.objmode(t0='f8'):
t0 = time.time()
#nthr = numba.get_num_threads()
s = np.zeros((nthr, X.shape[1]))
ss = np.zeros((nthr, X.shape[1]))
mean = np.zeros(X.shape[1])
std = np.zeros(X.shape[1])
n = X.shape[0]
for i in numba.prange(nthr):
for r in range(i, n, nthr):
for c in range(X.shape[1]):
v = X[r, c]
s[i, c] += v
ss[i, c] += v * v
for c in numba.prange(X.shape[1]):
s0 = s[:, c].sum()
mean[c] = s0 / n
std[c] = np.sqrt((ss[:, c].sum() - s0 * s0 / n) / (n - 1))
with numba.objmode():
print("finshed getting means, stddev", time.time() - t0)
for (r, c) in numba.pndindex(X.shape):
v = (X[r, c] - mean[c]) / std[c]
X[r, c] = maxv if v > maxv else v
with numba.objmode():
print("finshed scaling values", time.time() - t0)
def normal_cone_angles(normals: np.ndarray, threshold: float, min_norm):
assert normals.ndim == 4
size = normals.shape[0]
assert normals.shape == (size, size, size, 3)
assert size > 2
result = np.zeros((size - 2, size - 2, size - 2), dtype=np.bool8)
for i in numba.pndindex((size - 2, size - 2, size - 2)):
# Collect normals for position i
current = np.empty((26, 3), dtype=np.float32) # 26 possible neighbors
ci: numba.uint32 = 0
for o in np.ndindex((3, 3, 3)):
if o != (1, 1, 1):
x, y, z = i[0] + o[0], i[1] + o[1], i[2] + o[2]
value = normals[x, y, z]
norm = np.linalg.norm(value)
if norm > min_norm: # Only add if norm is valid
current[ci] = value / norm
ci += 1
if ci > 2:
valid = current[:ci]
# Check angle between all valid normals
result[i[0], i[1],
i[2]] = np.any(np.arccos(valid @ valid.T) > threshold)
return result
def set_array_normals(target: np.ndarray, mask: np.ndarray,
values: np.ndarray):
assert target.shape == mask.shape and mask.shape == values.shape
for i in numba.pndindex(target.shape):
x, y, z = i
if mask[x, y, z]:
target[x, y, z] = values[x, y, z]
def _collapse_dz(deltas: np.ndarray):
col = np.zeros(deltas.shape, dtype=deltas.dtype)
for xy in numba.pndindex((deltas.shape[0], deltas.shape[1])):
x, y = xy
z0 = 0
value0 = np.nan
d0 = 1 if value0 > 0 else -1
for z in range(deltas.shape[2]):
current = deltas[x, y, z]
if current and value0 == current:
col[x, y, z0] += d0
continue
z0 = z
value0 = current
d0 = 1 if value0 > 0 else -1
if current:
col[x, y, z0] = d0
return col
def _collapse_dx(deltas: np.ndarray):
col = np.zeros(deltas.shape, dtype=deltas.dtype)
for yz in numba.pndindex((deltas.shape[1], deltas.shape[2])):
y, z = yz
x0 = 0
value0 = np.nan
d0 = 1 if value0 > 0 else -1
for x in range(deltas.shape[0]):
current = deltas[x, y, z]
if current and value0 == current:
col[x0, y, z] += d0
continue
x0 = x
value0 = current
d0 = 1 if value0 > 0 else -1
if current:
col[x0, y, z] = d0
return col
def _sum(X):
s = 0
for i in numba.pndindex(X.shape):
s += X[i]
return s