def gpoints(SceneArea, npoints, amax=1., seed=None):
"""Generates point targets
Parameters
----------
SceneArea : tulpe or list
The area of scene.
npoints : int
The number of points.
amax : float, optional
The maximum amplitude.
seed : int or None, optional
The seed for random generator.
Returns
-------
tensor
A tensor contains coordinate and amplitude information.
"""
if seed is not None:
setseed(seed)
xmin, xmax, ymin, ymax = SceneArea
xs = th.rand(npoints, 1) * (xmax - xmin) + xmin
ys = th.rand(npoints, 1) * (ymax - ymin) + ymin
amps = th.rand(npoints, 1) * amax
targets = th.zeros(npoints, 7)
targets[:, [0]] = xs
targets[:, [1]] = ys
targets[:, [-1]] = amps
return targets
def mkpec(self, n, seed=None):
"""Make phase error
Args:
n (int): The number of phase errors.
seed (None or int, optional): The random seed.
Returns:
TYPE: Description
"""
if seed is not None:
ts.setseed(seed)
if self.ma is not None:
ca = th.zeros(n, self.ma - 1)
for i, lowhigh in enumerate(zip(self.carange[0], self.carange[1])):
low, high = lowhigh
ca[:, i] = th.rand(n) * (high - low) + low
if self.mr is not None:
cr = th.zeros(n, self.mr - 1)
for i, lowhigh in enumerate(zip(self.crrange[0], self.crrange[1])):
low, high = lowhigh
cr[:, i] = th.rand(n) * (high - low) + low
if self.ma is not None and self.mr is None:
return ca
if self.mr is not None and self.ma is None:
return cr
if self.ma is not None and self.mr is not None:
return ca, cr
def mksarp(self, n=1, seed=None):
"""Makes SAR Parameters
Parameters
----------
n : int, optional
The number of experiments.
seed : None, optional
The seed for random generator.
Returns
-------
dict
The SAR Parameter dict.
"""
if seed is not None:
ts.setseed(seed)
pdict = {}
for k, v in self.prdict.items():
if v is not None:
low, high = v
if n == 1:
pdict[k] = th.rand(n).item() * (high - low) + low
else:
pdict[k] = th.rand(n) * (high - low) + low
return pdict
def __init__(self, carange=None, crrange=None, seed=None):
"""Polynominal phase error generator.
Args:
carange (None or tuple or list, optional): List of coefficients range of phase error in azimuth direction.
crrange (None or tuple or list, optional): List of coefficients range of phase error in range direction.
seed (None or int, optional): The random seed.
"""
self.carange = carange
self.crrange = crrange
self.seed = seed
self.ma = len(
carange[0]) + 1 if carange is not None else None # order 2-ma
self.mr = len(
crrange[0]) + 1 if crrange is not None else None # order 2-mr
if seed is not None:
ts.setseed(seed)
if self.std is None:
if self.axis is None:
std = th.std(x, self.unbiased)
else:
std = th.std(x, self.axis, self.unbiased, keepdim=True)
if self.extra is True:
return (x - mean) / (std + EPS), mean, std
else:
return (x - mean) / (std + EPS)
if __name__ == '__main__':
import torchsar as ts
ts.setseed(seed=2020, target='torch')
x = th.randn(5, 2, 4, 3) * 10000
f = Standardization(axis=(2, 3), unbiased=False, extra=True)
y, meanv, stdv = f(x)
print(y[0], y.shape, meanv.shape, stdv.shape)
g = th.nn.InstanceNorm2d(2)
z = g(x)
print(z[0], z.shape)
f = Standardization(axis=(0, 2, 3), unbiased=False, extra=True)
y, meanv, stdv = f(x)
print(y[0], y.shape, meanv.shape, stdv.shape)
def __init__(self, prdict, seed=None):
self.prdict = prdict
self.seed = seed
if seed is not None:
ts.setseed(seed)