deepmie/make_dataset.py

import snlay as snlay
import h5py
import numpy as np
from scattnlay import scattnlay
import time
import progressbar


dataset_size = 100000 

# parameters of the dataset
num_layers = 8
num_lpoints = 250
lam_min = 300
lam_max = 1200
size_min = 30
size_max = 70

#make the materials list here
mats = [3, 4, 3, 4, 3, 4, 3, 4]       # 2 - silicon, silicon.dat,   1 - gold,  gold.dat
                                      # 3 - silica,  silica.dat,    4 - titania,  tio2.dat
#generate a huge array and then reshape
dataset_X = np.random.randint(size_min,size_max+1,num_layers*dataset_size).astype(float).reshape(dataset_size, num_layers)
lams = np.linspace(lam_min, lam_max, num_lpoints)
dataset_Y = np.zeros((dataset_size,num_lpoints))

for ind in progressbar.progressbar(np.arange(dataset_size)):
    kr, m = snlay.make_xm(dataset_X[ind,:], mats, lams)
    terms, dataset_Y[ind,:], Qsca, Qabs, Qbk, Qpr, g, Albedo, S1, S2 = scattnlay(kr, m)


h5f = h5py.File('./datasets/s8_sio2tio2.h5', 'w')
h5f.create_dataset('sizes', data=dataset_X)
h5f.create_dataset('spectrum', data=dataset_Y)
h5f.close()