def main(dataset_path,save_address,test_flag,verbose):
# Default input dataset path
dataset_path += '/'
# default address to save pickle files
save_address += '/'
# if input dataset is the collection of unlabelled glitches it is "1",
# if it is golden set set, it is "0".
if verbose:
print 'input dataset path', dataset_path
print 'save adress', save_address
print 'test flag', test_flag
if not os.path.exists(save_address):
if verbose:
print ('making... ' + save_address)
os.makedirs(save_address)
# these lines read all images with duration 1.0, 0.5, 2.0 and 4 seconds
imgs_1, label1, name1 = my_read_image(dataset_path, 'spectrogram_1.0.png')
imgs_1_fl = np.array(imgs_1, dtype='f')
"shuffling indexes for splitting data into training, validation, and test"
data_size = imgs_1_fl.shape[0]
new_ind = np.random.permutation(range(data_size))
imgs_5, label5, name5 = my_read_image(dataset_path, 'spectrogram_0.5.png')
imgs_5_fl = np.array(imgs_5, dtype='f')
imgs_2, label2, name2 = my_read_image(dataset_path, 'spectrogram_2.0.png')
imgs_2_fl = np.array(imgs_2, dtype='f')
imgs_4, label4, name4 = my_read_image(dataset_path, 'spectrogram_4.0.png')
imgs_4_fl = np.array(imgs_4, dtype='f')
# check if all four durations for each image exist
er1 = set(name1) ^ set(name2)
er2 = set(name2) ^ set(name4)
er3 = set(name5) ^ set(name4)
if (len(er1)):
print ('er1=', er1)
sys.exit('Error!! these images do not have all 4 durations')
if (len(er2)) :
print ('er2=', er2)
sys.exit('Error!! these images do not have all 4 durations')
if (len(er3)):
print ('er3=', er3)
sys.exit('Error!! these images do not have all 4 durations')
# writing into pickle files
if test_flag:
# for unlabelled glitches
my_save_dataset_test(save_address, imgs_1_fl, label1, name1, '_1.0_')
my_save_dataset_test(save_address, imgs_4_fl, label4, name4, '_4.0_')
my_save_dataset_test(save_address, imgs_2_fl, label2, name2, '_2.0_')
my_save_dataset_test(save_address, imgs_5_fl, label5, name5, '_5.0_')
# for golden set
else:
my_save_dataset(save_address, new_ind, imgs_1_fl, label1, name1, '_1.0_')
my_save_dataset(save_address, new_ind, imgs_4_fl, label4, name4, '_4.0_')
my_save_dataset(save_address, new_ind, imgs_2_fl, label2, name2, '_2.0_')
my_save_dataset(save_address, new_ind, imgs_5_fl, label5, name5, '_5.0_')
def main(dataset_path,save_address,test_flag,verbose):
np.random.seed(1986) # for reproducibility
# Default input dataset path
dataset_path += '/'
# default address to save pickle files
save_address += '/'
# if input dataset is the collection of unlabelled glitches it is "1",
# if it is golden set set, it is "0".
if verbose:
print 'input dataset path', dataset_path
print 'save adress', save_address
print 'test flag', test_flag
if not os.path.exists(save_address):
if verbose:
print ('making... ' + save_address)
os.makedirs(save_address)
# these lines read all images with duration 1.0, 0.5, 2.0 and 4 seconds
imgs_1, label1, name1, classes = my_read_image2(dataset_path, 'spectrogram_1.0.png')
imgs_1_fl = np.array(imgs_1, dtype='f')
imgs_2 = []
label2 = []
name2 = []
imgs_4 = []
label4 = []
name4 = []
imgs_5 = []
label5 = []
name5 = []
# for index, cl in enumerate(classes):
for idx, fi in enumerate(name1):
path5 = dataset_path + '/' + classes[label1[idx]] + '/' + fi + '_spectrogram_0.5.png'
path2 = dataset_path + '/' + classes[label1[idx]] + '/' + fi + '_spectrogram_2.0.png'
path4 = dataset_path + '/' + classes[label1[idx]] + '/' + fi + '_spectrogram_4.0.png'
file5 = io.imread(path5)
file2 = io.imread(path2)
file4 = io.imread(path4)
file5 = file5[66:532, 105:671, :]
file2 = file2[66:532, 105:671, :]
file4 = file4[66:532, 105:671, :]
#plt.imshow(file5)
file5 = rgb2gray(file5)
file2 = rgb2gray(file2)
file4 = rgb2gray(file4)
file5 = rescale(file5, [0.1, 0.1])
file2 = rescale(file2, [0.1, 0.1])
file4 = rescale(file4, [0.1, 0.1])
dim = np.int(reduce(lambda x, y: x * y, file5.shape))
imgs_5.append(np.reshape(file5, (dim)))
imgs_2.append(np.reshape(file2, (dim)))
imgs_4.append(np.reshape(file4, (dim)))
'''name5.append(fi)
name2.append(fi)
name4.append(fi)
label5.append(index)
label2.append(index)
label4.append(index) '''
name2 = name1
name4 = name1
name5 = name1
label2 = label1
label4 = label1
label5 = label1
# imgs_5, label5, name5 = my_read_image(dataset_path, 'spectrogram_0.5.png')
imgs_5_fl = np.array(imgs_5, dtype='f')
# imgs_2, label2, name2 = my_read_image(dataset_path, 'spectrogram_2.0.png')
imgs_2_fl = np.array(imgs_2, dtype='f')
# imgs_4, label4, name4 = my_read_image(dataset_path, 'spectrogram_4.0.png')
imgs_4_fl = np.array(imgs_4, dtype='f')
"shuffling indexes for splitting data into training, validation, and test"
data_size = imgs_1_fl.shape[0]
new_ind = np.random.permutation(range(data_size))
'''assert (name1 == name2)
assert (name2 == name4)
assert (name4 == name5)
# check if all four durations for each image exist
er1 = set(name1) ^ set(name2)
er2 = set(name2) ^ set(name4)
er3 = set(name5) ^ set(name4)
if (len(er1)):
print ('er1=', er1)
sys.exit('Error!! these images do not have all 4 durations')
if (len(er2)) :
print ('er2=', er2)
sys.exit('Error!! these images do not have all 4 durations')
if (len(er3)):
print ('er3=', er3)
sys.exit('Error!! these images do not have all 4 durations') '''
# writing into pickle files
if test_flag:
# for unlabelled glitches
my_save_dataset_test(save_address, imgs_1_fl, label1, name1, '_1.0_')
my_save_dataset_test(save_address, imgs_4_fl, label4, name4, '_4.0_')
my_save_dataset_test(save_address, imgs_2_fl, label2, name2, '_2.0_')
my_save_dataset_test(save_address, imgs_5_fl, label5, name5, '_5.0_')
else:
# for golden set
my_save_dataset(save_address, new_ind, imgs_1_fl, label1, name1, '_1.0_')
my_save_dataset(save_address, new_ind, imgs_4_fl, label4, name4, '_4.0_')
my_save_dataset(save_address, new_ind, imgs_2_fl, label2, name2, '_2.0_')
my_save_dataset(save_address, new_ind, imgs_5_fl, label5, name5, '_5.0_')
