def __init__(self, filter_size: int, xextent: tuple, yextent: tuple, img_shape: tuple):
"""
Args:
filter_size (int): size of the mean filter
xextent (tuple): extent in x
yextent (tuple): extent in y
img_shape (tuple): image shape
"""
super().__init__()
from decode.neuralfitter.utils import padding_calc as padcalc
"""Sanity checks"""
if filter_size % 2 == 0:
raise ValueError("ROI size must be odd.")
self.filter_size = [filter_size, filter_size]
self.img_shape = img_shape
pad_x = padcalc.pad_same_calc(self.img_shape[0], self.filter_size[0], 1, 1)
pad_y = padcalc.pad_same_calc(self.img_shape[1], self.filter_size[1], 1, 1)
self.padding = torch.nn.ReplicationPad2d((pad_x, pad_x, pad_y, pad_y)) # to get the same output dim
self.kernel = torch.ones((1, 1, filter_size, filter_size)) / (filter_size * filter_size)
self.delta_psf = psf_kernel.DeltaPSF(xextent, yextent, img_shape)
self.bin_x = self.delta_psf._bin_x
self.bin_y = self.delta_psf._bin_y
def __init__(self, xextent, yextent, img_shape):
super().__init__(ix_low=0, ix_high=0)
self.xextent = xextent
self.yextent = yextent
self.img_shape = img_shape
self.delta = psf_kernel.DeltaPSF(xextent=self.xextent,
yextent=self.yextent,
img_shape=self.img_shape)
def __init__(self,
*,
xextent: tuple,
yextent: tuple,
img_shape: tuple,
roi_size: int,
weight_mode='const',
weight_power: float = None,
forward_safety: bool = True,
ix_low: Union[int, None] = None,
ix_high: Union[int, None] = None,
squeeze_batch_dim: bool = False):
"""
Args:
xextent (tuple): extent in x
yextent (tuple): extent in y
img_shape: image shape
roi_size (int): roi size of the target
weight_mode (str): constant or phot
weight_power (float): power factor of the weight
forward_safety: check sanity of forward arguments
"""
super().__init__(ix_low=ix_low,
ix_high=ix_high,
squeeze_batch_dim=squeeze_batch_dim)
self.roi_size = roi_size
self.target_equivalent = target_generator.UnifiedEmbeddingTarget(
xextent=xextent,
yextent=yextent,
img_shape=img_shape,
roi_size=roi_size,
ix_low=ix_low,
ix_high=ix_high)
self.weight_psf = psf_kernel.DeltaPSF(xextent, yextent, img_shape)
self.weight_mode = weight_mode
self.weight_power = weight_power if weight_power is not None else 1.0
self._forward_safety = forward_safety
self.check_sanity()
def delta_05px(self):
return psf_kernel.DeltaPSF(xextent=(-0.5, 31.5),
yextent=(-0.5, 31.5),
img_shape=(64, 64))
def delta_1px(self):
return psf_kernel.DeltaPSF(xextent=(-0.5, 31.5),
yextent=(-0.5, 31.5),
img_shape=(32, 32))