def gen_rep_img(isABP):
if isABP:
int_dat = os.listdir('10 ABP int')
else:
int_dat = os.listdir('10 ICP int')
file_names = []
total_sig = []
total_lab = []
total_tim = []
for int_d in int_dat:
if isABP:
sig, lab, tim = ae.load_one_data('10 ABP int/' + int_d)
else:
sig, lab, tim = ae.load_one_data('10 ICP int/' + int_d)
total_sig.append(sig)
total_lab.append(lab)
total_tim.append(tim)
file_names.append(int_d)
for i in range(len(total_sig)):
signal_tag = 'ABP' if isABP else 'ICP'
if os.path.isfile('npy10/' + signal_tag + '/' + file_names[i] +
'.train.ori.npy'):
continue
print(i)
train_x = []
train_y = []
train_t = []
test_x = []
test_y = []
test_t = []
for j in range(len(total_sig)):
if i == j:
test_x.extend(total_sig[j])
test_y.extend(total_lab[j])
test_t.extend(total_tim[j])
else:
train_x.extend(total_sig[j])
train_y.extend(total_lab[j])
train_t.extend(total_tim[j])
train_x = ae.signal_to_img(train_x)
test_x = ae.signal_to_img(test_x)
train_rep_x = ae.autoencoding_cnn(
train_x, train_x, 'net/' + str(i) + '_' + signal_tag + '.net')
test_rep_x = ae.autoencding_cnn_using_net(
test_x, 'net/' + str(i) + '_' + signal_tag + '.net')
np.save('npy10/' + signal_tag + '/' + file_names[i] + '.train.ori.npy',
train_x)
np.save('npy10/' + signal_tag + '/' + file_names[i] + '.train.rep.npy',
train_rep_x)
np.save('npy10/' + signal_tag + '/' + file_names[i] + '.train.etc.npy',
[train_y, train_t])
np.save('npy10/' + signal_tag + '/' + file_names[i] + '.test.ori.npy',
test_x)
np.save('npy10/' + signal_tag + '/' + file_names[i] + '.test.rep.npy',
test_rep_x)
np.save('npy10/' + signal_tag + '/' + file_names[i] + '.test.etc.npy',
[test_y, test_t])
def remove_artifact(isABP, int_path):
samples = os.listdir(int_path)
signal_tag = 'ABP' if isABP else 'ICP'
rep_model = load_model('net/FINAL_' + signal_tag + '.net')
cla_model = load_model('net/CNN/FINAL_' + signal_tag + '.net')
for v in samples:
sig, lab, tim = ae.load_one_data(int_path + '/' + v)
train_x = ae.signal_to_img(train_x)
train_rep_x = ae.autoencding_cnn_using_net(train_x, rep_model)
def remove_label_generator(isABP, int_path):
signal_tag = 'ABP' if isABP else 'ICP'
model = load_model('net/CNN/FINAL_' + signal_tag + '.net')
files = os.listdir(int_path)
for file in files:
if file.split('.')[-1] != 'csv': continue
print(file)
sig, lab, tim = ae.load_one_data(int_path + '/' + file)
img = ae.signal_to_img(sig)
rep = ae.autoencding_cnn_using_net(img,
'net/FINAL_' + signal_tag + '.net')
pred = model.predict(rep)
pen = open('result/' + signal_tag + '/' + file + '.prd.fin.csv', 'w')
for i in range(len(pred)):
pen.write(str(i) + ',' + str(round(pred[i][0])) + '\n')
pen.close()
print('fin')
def gen_full_train_network(isABP):
signal_tag = 'ABP' if isABP else 'ICP'
int_dat = os.listdir('10 ' + signal_tag + ' int')
train_x = []
train_y = []
train_t = []
for int_d in int_dat:
sig, lab, tim = ae.load_one_data('10 ' + signal_tag + ' int/' + int_d)
train_x.extend(sig)
train_y.extend(lab)
train_t.extend(tim)
train_x = ae.signal_to_img(train_x)
train_rep = ae.autoencoding_cnn(train_x, train_x,
'net/FINAL_' + signal_tag + '.net')
new_train = n_augmentation(train_rep, label_gen(train_y), train_t)
model = trainer.create_model(64)
print(model.summary())
model.fit(np.array(new_train[0]),
np.array(new_train[1]),
epochs=10,
shuffle=True)
model.save('net/CNN/FINAL_' + signal_tag + '.net')
def make_img():
import matplotlib.pyplot as plt
int_dat = os.listdir('10 ICP int')
for int_d in int_dat:
sig, lab, tim = ae.load_one_data('10 ICP int/' + int_d)
img = ae.signal_to_img(sig)
if not os.path.isdir('img/' + int_d.split('_')[0]):
os.mkdir('img/' + int_d.split('_')[0])
os.mkdir('img/' + int_d.split('_')[0] + '/nor')
os.mkdir('img/' + int_d.split('_')[0] + '/art')
else:
continue
for i in range(len(tim)):
plt.figure(1)
plt.imshow(img[i].reshape(64, 64))
if lab[i] == 0:
plt.savefig('img/' + int_d.split('_')[0] + '/nor/' + str(i) +
'.png')
else:
plt.savefig('img/' + int_d.split('_')[0] + '/art/' + str(i) +
'.png')
plt.cla()
plt.clf()
def representation():
s, f = ae.load_data('10 ABP int/')
total_image = []
total_signal = []
file_len_map = {}
bef_cnt = 0
for k in range(len(f)):
filename = f[k].split('_')[0]
imgs = ae.signal_to_img(s[k])
total_image.extend(imgs)
total_signal.extend(s[k])
file_len_map[filename] = [bef_cnt, bef_cnt + len(imgs)]
bef_cnt += len(imgs)
total_rep_imgs = ae.autoencoding_cnn(total_image, total_image, fold='tot')
for l, v in file_len_map.items():
npy_ori = []
npy_rep = []
for m in range(v[0], v[1]):
npy_ori.append(np.array(total_image[m]).reshape(64, 64, 1))
npy_rep.append(np.array(total_rep_imgs[m]).reshape(64, 64, 1))
np.save('npy/abp/tot/' + l + '.ori.npy', np.array(npy_ori))
np.save('npy/abp/tot/' + l + '.rep.npy', np.array(npy_rep))
if not os.path.exists('img/ori/'+f.split('_'[0])):
os.mkdir('img/ori/'+f.split('_')[0])
if not os.path.exists('img/rep/'+f.split('_'[0])):
os.mkdir('img/rep/'+f.split('_')[0])
for i in range(len(lab)):
cur_img = ae.signal_to_img(sig[i])
"""
signals, filenames = ae.load_data(path)
total_image = []
total_signal = []
file_len_map = {}
bef_cnt = 0
for i in range(len(filenames)):
filename = filenames[i].split('_')[0]
imgs = ae.signal_to_img(signals[i])
total_image.extend(imgs)
total_signal.extend(signals[i])
file_len_map[filename] = [bef_cnt, bef_cnt + len(imgs)]
bef_cnt += len(imgs)
#total_rep_imgs = ae.autoencoding_cnn(total_image, total_image, img_dim=64, encoding_dim=16)
for k, v in file_len_map.items():
npy_arr = []
for i in range(v[0], v[1]):
plt.figure(1)
plt.imshow(total_image[i].reshape(64, 64))
plt.savefig('img/ori_b/' + k + '_' + str(i - v[0]) + '.png')
plt.cla()
plt.clf()
for j in range(5):
if i == j:
sa.extend(signals[j])
fa.extend(filenams[j])
else:
s.extend(signals[j])
f.extend(filenams[j])
total_image = []
total_signal = []
file_len_map = {}
bef_cnt = 0
for k in range(len(f)):
filename = f[k].split('_')[0]
imgs = ae.signal_to_img(s[k])
total_image.extend(imgs)
total_signal.extend(s[k])
file_len_map[filename] = [bef_cnt, bef_cnt + len(imgs)]
bef_cnt += len(imgs)
total_rep_imgs = ae.autoencoding_cnn(total_image, total_image, fold=i)
for l, v in file_len_map.items():
npy_ori = []
npy_rep = []
for m in range(v[0], v[1]):
npy_ori.append(total_image[m])
npy_rep.append(total_rep_imgs[m])
np.save('npy/' + str(i) + '/' + l + '.abp.ori.npy', npy_ori)
np.save('npy/' + str(i) + '/' + l + '.abp.rep.npy', npy_rep)