histogram#

class diffsptk.Histogram(n_bin=10, lower_bound=0, upper_bound=1, norm=False, softness=0.001)[source]#

See this page for details.

Parameters:

n_binint >= 1: Number of bins, $K$ .
lower_boundfloat < U: Lower bound of the histogram, $L$ .
upper_boundfloat > L: Upper bound of the histogram, $U$ .
normbool: If True, normalize the histogram.
softnessfloat > 0: A smoothing parameter. The smaller value makes the output closer to the true histogram, but the gradient vanishes.

References

[1]

M. Avi-Aharon et al., “DeepHist: Differentiable joint and color histogram layers for image-to-image translation,” arXiv preprint arXiv:2005.03995, 2020.

forward(x)[source]#

Compute histogram.

Parameters:

xTensor [shape=(…, T)]: Input data.

Returns:

outTensor [shape=(…, K)]: Histogram.

Examples

>>> x = diffsptk.ramp(9)
>>> histogram = diffsptk.Histogram(n_bin=4, lower_bound=-0.1, upper_bound=9.1)
>>> h = histogram(x)
>>> h
tensor([3., 2., 2., 3.])

diffsptk.functional.histogram(x, n_bin=10, lower_bound=0, upper_bound=1, norm=False, softness=0.001)[source]#

Compute histogram.

Parameters:

xTensor [shape=(…, T)]: Input data.
n_binint >= 1: Number of bins, $K$ .
lower_boundfloat < U: Lower bound of the histogram, $L$ .
upper_boundfloat > L: Upper bound of the histogram, $U$ .
normbool: If True, normalize the histogram.
softnessfloat > 0: A smoothing parameter. The smaller value makes the output closer to the true histogram, but the gradient vanishes.

Returns:

outTensor [shape=(…, K)]: Histogram in [L, U].