Created
November 25, 2021 04:49
-
-
Save Totoro97/88bd7f7adf5978e574f7498c08d5cd17 to your computer and use it in GitHub Desktop.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
from matplotlib import pyplot as plt | |
import numpy as np | |
n_samples = 128 | |
# pdf of logistic distribution | |
def pdf(x, std): | |
return 0.25 / std / np.cosh(0.5 * x / std)**2 | |
# cdf of logistic distribution | |
def cdf(x, std): | |
return 1 / (1 + np.exp(-x / std)) | |
def calc_density(f, dfdt, std): | |
return (-pdf(f, std) * dfdt / cdf(f, std)).clip(0, 1e7) | |
def calc_weights(f, dfdt, section_len, std): | |
# direct weight construction | |
weight_direct = pdf(f, std) | |
weight_direct = weight_direct / np.sum(weight_direct) | |
# weight construction of NeuS | |
density = calc_density(f, dfdt, std) | |
alpha = 1 - np.exp(-density * section_len) | |
trans = np.cumprod(1 - alpha) | |
weight_neus = np.concatenate([alpha[:1], trans[:-1] * alpha[1:]]) | |
return weight_direct, weight_neus | |
case = 'single_plane' # or 'multiple surfaces' | |
if __name__ == '__main__': | |
section_len = 1.0 / n_samples * np.ones(n_samples) | |
steps = np.linspace(0, 1, n_samples + 1)[:-1] | |
if case == 'single_plane': | |
f = np.linspace(0.5, -0.5, n_samples + 1)[:-1] # signed distance function | |
dfdt = -np.ones(n_samples) # df/dt | |
else: | |
f = np.concatenate([ | |
np.linspace(0.125, -0.125, n_samples // 4 + 1)[:-1], | |
np.linspace(-0.125, 0.125, n_samples // 4 + 1)[:-1], | |
np.linspace(0.125, -0.125, n_samples // 4 + 1)[:-1], | |
np.linspace(-0.125, 0.125, n_samples // 4 + 1)[:-1]]) | |
dfdt = np.concatenate([ | |
-np.ones(n_samples // 4), | |
np.ones(n_samples // 4), | |
-np.ones(n_samples // 4), | |
np.ones(n_samples // 4) | |
]) | |
for s in range(1, 20): | |
weight_direct, weight_neus = calc_weights(f, dfdt, section_len, 1 / (1.5**s)) | |
plt.plot(steps, weight_direct) | |
plt.plot(steps, weight_neus) | |
scaled_f = f / f.max() * np.max([weight_direct.max(), weight_neus.max()]) | |
plt.plot(steps, scaled_f) | |
plt.savefig('./tmp_plot_single_plane/{}.png'.format(s)) | |
plt.cla() | |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment