def __call__(self, colat_idx, lon_idx, f, r):
"""
Interpolate function samples at order `N`.
Parameters
----------
colat_idx : :py:class:`~numpy.ndarray`
(N_colat,) polar support indices from :py:func:`~imot_tools.math.sphere.grid.equal_angle`.
lon_idx : :py:class:`~numpy.ndarray`
(N_lon,) azimuthal support indices from :py:func:`~imot_tools.math.sphere.grid.equal_angle`.
f : :py:class:`~numpy.ndarray`
(L, N_colat, N_lon) zonal function values at support points. (float or complex)
r : :py:class:`~numpy.ndarray`
(3, N_px) evaluation points.
Returns
-------
f_interp : :py:class:`~numpy.ndarray`
(L, N_px) function values at specified coordinates.
"""
N_colat = colat_idx.size
if not (colat_idx.shape == (N_colat, )):
raise ValueError(
"Parameter[colat_idx] must have shape (N_colat,).")
N_lon = lon_idx.size
if not (lon_idx.shape == (N_lon, )):
raise ValueError("Parameter[lon_idx] must have shape (N_lon,).")
L = len(f)
if not (f.shape == (L, N_colat, N_lon)):
raise ValueError(
"Parameter[f] must have shape (L, N_colat, N_lon).")
if not ((r.ndim == 2) and (r.shape[0] == 3)):
raise ValueError("Parameter[r] must have shape (3, N_px).")
# Apply weights directly onto `f` to avoid memory blow-up.
_, _, colat, lon = grid.equal_angle(self._N)
a = np.arange(self._N + 1)
weight = (np.sum(np.sin((2 * a + 1) * colat[colat_idx]) / (2 * a + 1),
axis=1) * np.sin(colat[colat_idx, 0]) *
((2 * np.pi) / ((self._N + 1)**2))) # (N_colat,)
fw = f * weight.reshape((1, N_colat, 1)) # (L, N_colat, N_lon)
f_interp = super().__call__(
weight=np.broadcast_to([1], (N_colat * N_lon, )),
support=transform.pol2cart(1, colat[colat_idx, :],
lon[:, lon_idx]).reshape((3, -1)),
f=fw.reshape((L, -1)),
r=r,
sparsity_mask=None,
)
return f_interp
def spherical(direction, FoV, size):
"""
Spherical pixel grid.
Parameters
----------
direction : :py:class:`~numpy.ndarray`
(3,) vector around which the grid is centered.
FoV : float
Span of the grid [rad] centered at `direction`.
size : :py:class:`~numpy.ndarray`
(N_height, N_width)
The grid will consist of `N_height` concentric circles around `direction`, each containing
`N_width` pixels.
Returns
-------
XYZ : :py:class:`~numpy.ndarray`
(3, N_height, N_width) pixel grid.
"""
direction = np.array(direction, dtype=float)
direction /= linalg.norm(direction)
if not (0 < np.rad2deg(FoV) <= 179):
raise ValueError("Parameter[FoV] must be in (0, 179] degrees.")
size = np.array(size, copy=False)
if np.any(size <= 0):
raise ValueError("Parameter[size] must contain positive entries.")
N_height, N_width = size
colat, lon = np.meshgrid(np.linspace(0, FoV / 2, N_height),
np.linspace(0, 2 * np.pi, N_width),
indexing="ij")
XYZ = transform.pol2cart(1, colat, lon)
# Center grid at 'direction'
_, dir_colat, _ = transform.cart2pol(*direction)
R_axis = np.cross([0, 0, 1], direction)
if np.allclose(R_axis, 0):
# R_axis is in span(E_z), so we must manually set R
R = np.eye(3)
if direction[2] < 0:
R[2, 2] = -1
else:
R = ilinalg.rot(axis=R_axis, angle=dir_colat)
XYZ = np.tensordot(R, XYZ, axes=1)
return XYZ
def spherical_geometry():
"""
Spherical 3D microphone array geometry.
Radius: 0.20[m]
Returns
-------
XYZ : :py:class:`~numpy.ndarray`
(3, N_antenna) Cartesian coordinates.
"""
N_antenna = 64
n = np.arange(N_antenna)
colat = np.arccos(1 - (2 * n + 1) / N_antenna)
lon = (4 * np.pi * n) / (1 + np.sqrt(5))
r = 0.2
XYZ = np.stack(transform.pol2cart(r, colat, lon), axis=0)
return XYZ
def from_circular_distribution(direction, FoV, N_src):
"""
Distribute `N_src` sources on a circle centered at `direction`.
Parameters
----------
direction : :py:class:`~numpy.ndarray`
(3,) direction in the sky.
FoV : float
Spherical angle [rad] of the sky centered at `direction` from which sources are extracted.
N_src : int
Number of dominant sources to extract.
Returns
-------
sky_model : :py:class:`~deepwave.tools.data_gen.source.SkyModel`
Sky model.
"""
if not (0 < FoV < 2 * np.pi):
raise ValueError('Parameter[FoV] must lie in (0, 360) degrees.')
colat = FoV / 4
lon = np.linspace(0, 2 * np.pi, N_src, endpoint=False)
XYZ = np.stack(transform.pol2cart(1, colat, lon), axis=0)
# Center grid at 'direction'
_, dir_colat, _ = transform.cart2pol(*direction)
R_axis = np.cross([0, 0, 1], direction)
if np.allclose(R_axis, 0):
# R_axis is in span(E_z), so we must manually set R
R = np.eye(3)
if direction[2] < 0:
R[2, 2] = -1
else:
R = pylinalg.rot(axis=R_axis, angle=dir_colat)
XYZ = np.tensordot(R, XYZ, axes=1)
I = np.ones((N_src, ))
sky_model = SkyModel(XYZ, I)
return sky_model
def _from_fits(cls, primary_hdu, image_hdu):
"""
Load image from Header Descriptor Units.
Parameters
----------
primary_hdu : :py:class:`~astropy.io.fits.PrimaryHDU`
image_hdu : :py:class:`~astropy.io.fits.ImageHDU`
Returns
-------
I : :py:class:`~imot_tools.io.s2image.Image`
"""
# PrimaryHDU: grid specification.
colat, lon = primary_hdu.data
grid = transform.pol2cart(1, colat, lon)
# ImageHDU: extract data cube.
data = image_hdu.data
I = cls(data=data, grid=grid)
return I
def sky(f):
"""
Extract Sky Model from file name.
Parameters
==========
f : :py:class:`~pathlib.Path`
.wav file with name of the form "<characters>_spkr[012]_angle[0:360:2].wav"
Returns
=======
sky_model : :py:class:`~deepwave.tools.data_gen.source.SkyModel`
Container with source information.
`sky_model.intensity` will contain the speaker number {0, 1, 2}.
"""
pattern = r'.+_spkr(\d+)_angle(\d+)\.wav'
match = re.search(pattern, str(f))
if match:
idx_speaker, idx_angle = map(int, match.group(1, 2))
if idx_speaker not in (0, 1, 2):
raise ValueError(f'speaker number {idx_speaker} is out of bounds.')
if idx_angle not in np.arange(0, 360, 2):
raise ValueError(f'angle number {idx_angle} is out of bounds.')
m_speaker = speaker_map()
colat = data()['sources'][m_speaker[idx_speaker]]['colatitude'][str(
idx_angle)]
lon = data()['sources'][m_speaker[idx_speaker]]['azimuth'][str(
idx_angle)]
src_xyz = transform.pol2cart(1, colat, lon)
src_I = np.r_[idx_speaker]
sky_model = source.SkyModel(src_xyz, src_I)
return sky_model
else:
raise ValueError('Parameter[f] does not have the right form.')
def simulate_dataset(N_sample, N_src, XYZ, R, wl, src_mask, intensity=None, rate=None):
"""
Generate APGD dataset.
Parameters
----------
N_sample : int
Number of images to generate.
N_src : int
Number of sources present per image.
XYZ : :py:class:`~numpy.ndarray`
(3, N_antenna) microphone positions.
R : :py:class:`~numpy.ndarray`
(3, N_px) pixel grid.
wl : float
Wavelength [m] of plane wave.
src_mask : :py:class:`~numpy.ndarray`
(N_px,) boolean mask saying near which pixels it is possible to place sources.
intensity : float
If present, generate equi-amplitude sources.
rate : float
If present, generate rayleigh-amplitude sources.
Returns
-------
D : :py:class:`~acoustic_camera.nn.DataSet`
(N_sample,) dataset
Note
----
Either `intensity` or `rate` must be specified, not both.
"""
if not (N_sample > 0):
raise ValueError('Paremeter[N_sample] must be positive.')
if not (N_src > 0):
raise ValueError('Paremeter[N_src] must be positive.')
if not ((XYZ.ndim == 2) and (XYZ.shape[0] == 3)):
raise ValueError('Parameter[XYZ]: expected (3, N_antenna) array.')
N_antenna = XYZ.shape[1]
if not ((R.ndim == 2) and (R.shape[0] == 3)):
raise ValueError('Parameter[R]: expected (3, N_px) array.')
N_px = R.shape[1]
if wl < 0:
raise ValueError('Parameter[wl] is out of bounds.')
if not ((src_mask.ndim == 1) and (src_mask.size == N_px)):
raise ValueError('Parameter[src_mask]: expected (N_px,) boolean array.')
if (((intensity is None) and (rate is None)) or
((intensity is not None) and (rate is not None))):
raise ValueError('One of Parameters[intensity, rate] must be specified.')
if (intensity is not None) and (intensity <= 0):
raise ValueError('Parameter[intensity] must be positive.')
if (rate is not None) and (rate <= 0):
raise ValueError('Parameter[rate] must be positive.')
vis_gen = statistics.VisibilityGenerator(T=50e-3, fs=48000, SNR=10)
A = phased_array.steering_operator(XYZ, R, wl)
sampler = nn.Sampler(N_antenna, N_px)
N_data = sampler._N_cell
data = np.zeros((N_sample, N_data))
ground_truth = [None] * N_sample
apgd_gamma = 0.5
apgd_lambda_ = np.zeros(N_sample)
apgd_N_iter = np.zeros(N_sample)
apgd_tts = np.zeros(N_sample)
for i in range(N_sample):
logging.info(f'Generate APGD image {i + 1}/{N_sample}.')
### Create synthetic sky
if (intensity is not None):
sky_I = intensity * np.ones(N_src)
elif (rate is not None):
sky_I = stats.rayleigh.rvs(scale=rate, size=N_src)
sky_XYZ = R[:, src_mask][:, np.random.randint(0, np.sum(src_mask), size=N_src)]
## Randomize positions slightly to not fall straight onto grid.
_, sky_colat, sky_lon = transform.cart2pol(*sky_XYZ)
px_pitch = np.arccos(np.clip(R[:, 0] @ R[:, 1:], -1, 1)).min()
colat_noise, lon_noise = 0.1 * px_pitch * np.random.randn(2, N_src)
sky_XYZ = np.stack(transform.pol2cart(1, sky_colat + colat_noise, sky_lon + lon_noise), axis=0)
sky_model = source.SkyModel(sky_XYZ, sky_I)
S = vis_gen(XYZ, wl, sky_model)
# Normalize `S` spectrum for scale invariance.
S_D, S_V = linalg.eigh(S)
if S_D.max() <= 0:
S_D[:] = 0
else:
S_D = np.clip(S_D / S_D.max(), 0, None)
S = (S_V * S_D) @ S_V.conj().T
I_apgd = apgd.solve(S, A,
lambda_=None,
gamma=apgd_gamma,
L=None,
d=50,
x0=None,
eps=1e-3,
N_iter_max=200,
verbosity='NONE', # 'LOW',
momentum=True)
data[i] = sampler.encode(S=S, I=I_apgd['sol'])
ground_truth[i] = sky_model
apgd_lambda_[i] = I_apgd['lambda_']
apgd_N_iter[i] = I_apgd['niter']
apgd_tts[i] = I_apgd['time']
D = nn.DataSet(data, XYZ, R, wl, ground_truth,
apgd_lambda_, apgd_gamma, apgd_N_iter, apgd_tts)
return D
def fibonacci(N, direction=None, FoV=None):
r"""
(Region-limited) near-uniform sampling on the sphere.
Parameters
----------
N : int
Order of the grid, i.e. there will be :math:`4 (N + 1)^{2}` points on the sphere.
direction : :py:class:`~numpy.ndarray`
(3,) vector around which the grid is centered.
If :py:obj:`None`, then the grid covers the entire sphere.
FoV : float
Span of the grid [rad] centered at `direction`.
This parameter is ignored if `direction` left unspecified.
Returns
-------
XYZ : :py:class:`~numpy.ndarray`
(3, N_px) sample points.
`N_px == 4*(N+1)**2` if `direction` left unspecified.
Examples
--------
Sampling a zonal function :math:`f(r): \mathbb{S}^{2} \to \mathbb{C}` of order :math:`N` on the
sphere:
.. testsetup::
import numpy as np
from imot_tools.math.sphere.grid import fibonacci
.. doctest::
>>> N = 2
>>> XYZ = fibonacci(N)
>>> np.around(XYZ, 2)
array([[ 0.23, -0.29, 0.04, 0.36, -0.65, 0.61, -0.2 , -0.37, 0.8 ,
-0.81, 0.39, 0.28, -0.82, 0.95, -0.56, -0.13, 0.76, -1. ,
0.71, -0.05, -0.63, 0.97, -0.79, 0.21, 0.46, -0.87, 0.8 ,
-0.33, -0.27, 0.68, -0.7 , 0.36, 0.1 , -0.4 , 0.4 , -0.16],
[ 0. , -0.27, 0.51, -0.47, 0.12, 0.39, -0.74, 0.72, -0.29,
-0.34, 0.82, -0.89, 0.48, 0.21, -0.8 , 0.98, -0.64, -0.04,
0.71, -1. , 0.76, -0.13, -0.55, 0.93, -0.81, 0.28, 0.37,
-0.78, 0.76, -0.36, -0.18, 0.56, -0.58, 0.31, 0.03, -0.17],
[ 0.97, 0.92, 0.86, 0.81, 0.75, 0.69, 0.64, 0.58, 0.53,
0.47, 0.42, 0.36, 0.31, 0.25, 0.19, 0.14, 0.08, 0.03,
-0.03, -0.08, -0.14, -0.19, -0.25, -0.31, -0.36, -0.42, -0.47,
-0.53, -0.58, -0.64, -0.69, -0.75, -0.81, -0.86, -0.92, -0.97]])
Sampling a zonal function :math:`f(r): \mathbb{S}^{2} \to \mathbb{C}` of order :math:`N` on
*part* of the sphere:
.. doctest::
>>> N = 2
>>> direction = np.r_[1, 0, 0]
>>> FoV = np.deg2rad(90)
>>> XYZ = fibonacci(N, direction, FoV)
>>> np.around(XYZ, 2)
array([[ 0.8 , 0.95, 0.76, 0.71, 0.97, 0.8 ],
[-0.29, 0.21, -0.64, 0.71, -0.13, 0.37],
[ 0.53, 0.25, 0.08, -0.03, -0.19, -0.47]])
Notes
-----
The sample positions on the unit sphere are given (in radians) by [2]_:
.. math::
\cos(\theta_{q}) & = 1 - \frac{2 q + 1}{4 (N + 1)^{2}}, \qquad & q \in \{ 0, \ldots, 4 (N + 1)^{2} - 1 \},
\phi_{q} & = \frac{4 \pi}{1 + \sqrt{5}} q, \qquad & q \in \{ 0, \ldots, 4 (N + 1)^{2} - 1 \}.
.. [2] B. Rafaely, "Fundamentals of Spherical Array Processing", Springer 2015
"""
if direction is not None:
direction = np.array(direction, dtype=float)
direction /= linalg.norm(direction)
if FoV is not None:
if not (0 < np.rad2deg(FoV) < 360):
raise ValueError("Parameter[FoV] must be in (0, 360) degrees.")
else:
raise ValueError(
"Parameter[FoV] must be specified if Parameter[direction] provided."
)
if N < 0:
raise ValueError("Parameter[N] must be non-negative.")
N_px = 4 * (N + 1)**2
n = np.arange(N_px)
colat = np.arccos(1 - (2 * n + 1) / N_px)
lon = (4 * np.pi * n) / (1 + np.sqrt(5))
XYZ = np.stack(transform.pol2cart(1, colat, lon), axis=0)
if direction is not None: # region-limited case.
# TODO: highly inefficient to generate the grid this way!
min_similarity = np.cos(FoV / 2)
mask = (direction @ XYZ) >= min_similarity
XYZ = XYZ[:, mask]
return XYZ