def get_train_test(split_ratio=0.6, random_state=42):
labels, indices, _ = get_labels()
X = np.load(labels[0] + '.npy')
y = np.zeros(X.shape[0])
for i, label in enumerate(labels[1:]):
x = np.load(label + '.npy')
X = np.vstack((X, x))
y = np.append(y, np.full(x.shape[0], fill_value=(i + 1)))
assert X.shape[0] == len(y)
return train_test_split(X, y, test_size=(1 - split_ratio), random_state=random_state)
def main(out_file):
sys, dia = convert_data.get_labels(LABELS_FILE)
all_systole = np.array(sys.values())
all_diastole = np.array(dia.values())
default_sys = validate.get_default(all_systole)
default_dia = validate.get_default(all_diastole)
sys_net = validate.get_net(SYS_DEPLOY_FILE, SYS_MODEL_FILE)
dia_net = validate.get_net(DIA_DEPLOY_FILE, DIA_MODEL_FILE)
print 'Writing submission'
validate.write_nn_submission(
out_file, PROCESSED_VALIDATION_DATA_DIR,
sys_net, dia_net,
default_sys, default_dia,
PREPROC)
def process_all(train_dest_folder, test_dest_folder, data_folder,
labels_file, n_jobs=-1, use_lv_segmentation=False):
random.seed(SEED)
caffe.set_device(DEVICE_ID)
caffe.set_mode_gpu()
if labels_file:
systole, diastole = convert_data.get_labels(labels_file)
else:
systole = diastole = None
studies = sorted(convert_data.get_studies(data_folder))
test_prop = TEST_PROP if test_dest_folder else 0.0
dest_folders = ([test_dest_folder] * int(test_prop * len(studies)) +
[train_dest_folder] * (len(studies) - int(test_prop * len(studies))))
random.shuffle(dest_folders)
Parallel(n_jobs=n_jobs)(
delayed(_process)(dest_folder, study_id, study_path, systole, diastole, use_lv_segmentation)
for dest_folder, (study_id, study_path) in zip(dest_folders, studies))
# plt.draw()
# plt.pause(0.001)
m = np.abs(f).mean()
if m < 0:
print('-')
else:
inp = frames2mfcc(f).reshape(1, 20, 11, 1)
# print(get_labels()[0][
# np.argmax(model.predict(inp))
# ])
predicts = model.predict(inp)
if np.max(predicts) < 0.8:
print('.')
else:
print(get_labels()[0][np.argmax(predicts)],
np.max(predicts))
#print(len(frames))
except KeyboardInterrupt:
pass
# print("Terminating")
# stream.stop_stream()
# stream.close()
# audio.terminate()
# waveFile = wave.open(WAVE_OUTPUT_FILENAME, 'wb')
# waveFile.setnchannels(CHANNELS)
# waveFile.setsampwidth(audio.get_sample_size(FORMAT))
# waveFile.setframerate(RATE)
# waveFile.writeframes(b''.join(frames))
def submission():
"""
Generate submission file for the trained models.
"""
cdf_fn = vec_to_cdf
sys, dia = convert_data.get_labels(LABELS_FILE)
all_systole = np.array(sys.values())
all_diastole = np.array(dia.values())
default_sys = validate.get_default(all_systole)
default_dia = validate.get_default(all_diastole)
print('Loading and compiling models...')
model_systole = GET_MODEL(NUM_FRAMES, CROP_SIZE, CROP_SIZE, OUT)
model_diastole = GET_MODEL(NUM_FRAMES, CROP_SIZE, CROP_SIZE, OUT)
print('Loading models weights...')
model_systole.load_weights(SAVE_BEST_WEIGHTS_SYS)
model_diastole.load_weights(SAVE_BEST_WEIGHTS_DIA)
print('Loading validation data...')
X, ids = load_validation_data(X_VAL_FILE, IDS_FILE)
if PREPROCESS:
print('Pre-processing images...')
X = preprocess(X)
X = center(X, CROP_SIZE)
batch_size = 32
print('Predicting on validation data...')
pred_systole = model_systole.predict(X, batch_size=batch_size, verbose=1)
pred_diastole = model_diastole.predict(X, batch_size=batch_size, verbose=1)
print(pred_systole.shape)
print(pred_diastole.shape)
# real predictions to CDF
cdf_pred_systole = cdf_fn(pred_systole)
cdf_pred_diastole = cdf_fn(pred_diastole)
print(cdf_pred_systole.shape)
print(cdf_pred_diastole.shape)
print('Accumulating results...')
sub_systole = accumulate_study_results(ids, cdf_pred_systole)
sub_diastole = accumulate_study_results(ids, cdf_pred_diastole)
# write to submission file
print('Writing submission to file...')
fi = csv.reader(open(SAMPLE_SUBMISSION))
f = open(OUT_SUBMISSION, 'w')
fo = csv.writer(f, lineterminator='\n')
fo.writerow(fi.next())
for line in fi:
idx = line[0]
key, target = idx.split('_')
key = int(key)
out = [idx]
if key in sub_systole:
if target == 'Diastole':
out.extend(list(sub_diastole[key][0]))
else:
out.extend(list(sub_systole[key][0]))
else:
if target == 'Diastole':
out.extend(list(default_dia))
else:
out.extend(list(default_sys))
print('Miss {0}'.format(idx))
fo.writerow(out)
f.close()
print('Done.')
def addBackgroundNoise():
background, fs, enc = wavread(BACKGROUND_NOISE_FILE)
background = background[:,0]
labels, _, _ = get_labels(INPUT_DIRECTORY)
for label in labels:
generateSamples(label, background)
def submission():
"""
Generate submission file for the trained models.
"""
cdf_fn = vec_to_cdf
sys, dia = convert_data.get_labels(LABELS_FILE)
all_systole = np.array(sys.values())
all_diastole = np.array(dia.values())
default_sys = validate.get_default(all_systole)
default_dia = validate.get_default(all_diastole)
print('Loading and compiling models...')
model_systole = GET_MODEL(NUM_FRAMES, CROP_SIZE, CROP_SIZE, OUT)
model_diastole = GET_MODEL(NUM_FRAMES, CROP_SIZE, CROP_SIZE, OUT)
print('Loading models weights...')
model_systole.load_weights(SAVE_BEST_WEIGHTS_SYS)
model_diastole.load_weights(SAVE_BEST_WEIGHTS_DIA)
print('Loading validation data...')
X, ids = load_validation_data(X_VAL_FILE, IDS_FILE)
if PREPROCESS:
print('Pre-processing images...')
X = preprocess(X)
X = center(X, CROP_SIZE)
batch_size = 32
print('Predicting on validation data...')
pred_systole = model_systole.predict(X, batch_size=batch_size, verbose=1)
pred_diastole = model_diastole.predict(X, batch_size=batch_size, verbose=1)
print(pred_systole.shape)
print(pred_diastole.shape)
# real predictions to CDF
cdf_pred_systole = cdf_fn(pred_systole)
cdf_pred_diastole = cdf_fn(pred_diastole)
print(cdf_pred_systole.shape)
print(cdf_pred_diastole.shape)
print('Accumulating results...')
sub_systole = accumulate_study_results(ids, cdf_pred_systole)
sub_diastole = accumulate_study_results(ids, cdf_pred_diastole)
# write to submission file
print('Writing submission to file...')
fi = csv.reader(open(SAMPLE_SUBMISSION))
f = open(OUT_SUBMISSION, 'w')
fo = csv.writer(f, lineterminator='\n')
fo.writerow(fi.next())
for line in fi:
idx = line[0]
key, target = idx.split('_')
key = int(key)
out = [idx]
if key in sub_systole:
if target == 'Diastole':
out.extend(list(sub_diastole[key][0]))
else:
out.extend(list(sub_systole[key][0]))
else:
if target == 'Diastole':
out.extend(list(default_dia))
else:
out.extend(list(default_sys))
print('Miss {0}'.format(idx))
fo.writerow(out)
f.close()
print('Done.')
from keras.models import model_from_yaml
from mfcc import wav2mfcc
from convert_data import get_labels
import numpy as np
OUTPUT_PATH = './output/'
yaml_file = open(OUTPUT_PATH + 'model.yaml', 'r')
loaded_model_yaml = yaml_file.read()
yaml_file.close()
model = model_from_yaml(loaded_model_yaml)
model.load_weights(OUTPUT_PATH + 'model.h5')
# FIXME: think about model.compile if needed
# sample = wav2mfcc('./data/right/0a7c2a8d_nohash_0.wav')
sample = wav2mfcc('./recordings/1525038319.2531478.wav')
sample_reshaped = sample.reshape(1, 20, 11, 1)
print(get_labels()[0][np.argmax(model.predict(sample_reshaped))])