sample_images

radionets.evaluation.utils.sample_images(mean, std, num_samples, conf)

Samples for every pixel in Fourier space from a truncated Gaussian distribution based on the output of the network.

Parameters:

meantorch.tensor

mean values of the pixels with shape (number of images, number of samples, image size // 2 + 1, image_size)

stdtorch.tensor

uncertainty values of the pixels with shape (number of images, number of samples, image size // 2 + 1, image_size)

num_samplesint

number of samples in Fourier space

Returns:

dict

resulting mean and standard deviation