print("F1 score %.2f"%f)
file_=open('run1.txt','a')
str1="act3={}, input_neurons={}, batchsize={}, loss={}, nb_filter={}, filter_length={}, optimizer={}".format(act3, input_neurons, batchsize, loss,nb_filter, filter_length, optimizer)
str2="EER={0:.2f} Precision={1:.2f} Recall={2:.2f} F1 score={3:.2f}".format(eer,p,r,f)
file_.write(str1+'\n'+str2+'\n')
file_.close()
"""
#file_logger.close()
miz = aud_model.Functional_Model(input_neurons=input_neurons,
dropout=dropout,
num_classes=num_classes,
model=model,
dimx=dimx,
dimy=dimy,
nb_filter=nb_filter,
act1=act1,
act2=act2,
act3=act3,
filter_length=filter_length,
pool_size=pool_size,
optimizer=optimizer,
loss=loss)
#fit the model
#tr_y=to_categorical(tr_y,len(cfg.labels))
#v_y=to_categorical(v_y,len(cfg.labels))
#fold_='saved_models_3'
#if os.path.exists(fold_):
# os.rmdir(fold_)
#os.mkdir(fold_)
#filepath=fold_+"/weights-improvement1-{epoch:02d}-{val_acc:.2f}.hdf5"
#checkpoint = ModelCheckpoint(filepath, monitor='val_acc', verbose=1, save_best_only=True, mode='max')
#aud_utils.check_dimension(feature[i],dimy[i],'defaults.yaml')
#tr_X=aud_utils.equalise(tr_X)
for i in range(len(feature)):
tr_X[i]=aud_utils.mat_3d_to_nd(model,tr_X[i])
print(tr_X[i].shape)
dimx=tr_X[0].shape[-2]
if prep=='dev':
cross_validation=True
else:
cross_validation=False
miz=aud_model.Functional_Model(input_neurons=input_neurons,cross_validation=cross_validation,dropout1=dropout1,
act1=act1,act2=act2,act3=act3,nb_filter = nb_filter, filter_length=filter_length,
num_classes=num_classes,
model=model,dimx=dimx,dimy=dimy)
np.random.seed(68)
if cross_validation:
kf = KFold(len(tr_X[0]),folds,shuffle=True,random_state=42)
results=[]
for train_indices, test_indices in kf:
train_x = list(np.zeros(len(feature),dtype='int'))
test_x = list(np.zeros(len(feature),dtype='int'))
for i in range(len(feature)):
train_x[i] = [tr_X[i][ii] for ii in train_indices]
test_x[i] = [tr_X[i][ii] for ii in test_indices]
train_x[i] = np.array(train_x[i])
test_x[i] = np.array(test_x[i])
train_y = [tr_y[ii] for ii in train_indices]
print(tr_y.shape)
dimx = tr_X.shape[-2]
dimy = tr_X.shape[-1]
tr_X = aud_utils.mat_3d_to_nd(model, tr_X)
print(tr_X.shape)
if prep == 'dev':
cross_validation = True
else:
cross_validation = False
miz = aud_model.Functional_Model(input_neurons=input_neurons,
dropout=0.2,
num_classes=num_classes,
model=model,
dimx=dimx,
dimy=dimy,
loss=loss,
optimizer=optimizer)
np.random.seed(68)
if cross_validation:
kf = KFold(len(tr_X), folds, shuffle=True, random_state=42)
results = []
for train_indices, test_indices in kf:
train_x = [tr_X[ii] for ii in train_indices]
train_y = [tr_y[ii] for ii in train_indices]
test_x = [tr_X[ii] for ii in test_indices]
test_y = [tr_y[ii] for ii in test_indices]
train_y = to_categorical(train_y, num_classes=len(labels))
test_y = to_categorical(test_y, num_classes=len(labels))
print(tr_X.shape)
print(tr_y.shape)
tr_X = aud_utils.mat_3d_to_nd(model, tr_X)
print(tr_X.shape)
dimx = tr_X.shape[-2]
dimy = tr_X.shape[-1]
if prep == 'dev':
print "Number of folds", folds
cross_validation = True
else:
cross_validation = False
miz = aud_model.Functional_Model(num_classes=num_classes,
model=model,
dimx=dimx,
dimy=dimy)
np.random.seed(68)
if cross_validation:
kf = KFold(len(tr_X), folds, shuffle=True, random_state=42)
results = []
for train_indices, test_indices in kf:
train_x = [tr_X[ii] for ii in train_indices]
train_y = [tr_y[ii] for ii in train_indices]
test_x = [tr_X[ii] for ii in test_indices]
test_y = [tr_y[ii] for ii in test_indices]
#train_y = to_categorical(train_y,num_classes=len(labels))
#test_y = to_categorical(test_y,num_classes=len(labels))
train_x = np.array(train_x)
print(tr_X.shape)
dimx = tr_X.shape[-2]
dimy = tr_X.shape[-1]
if prep == 'dev':
print "Number of folds", folds
cross_validation = True
else:
cross_validation = False
rnn_units = [20, 20]
miz = aud_model.Functional_Model(num_classes=num_classes,
model=model,
dimx=dimx,
dimy=dimy,
dropout=dropout1,
filter_lenght=filter_length,
act1=act1,
act2=act2,
nb_filter=nb_filter,
rnn_units=rnn_units,
act3=act3)
np.random.seed(68)
if cross_validation:
kf = KFold(len(tr_X), folds, shuffle=True, random_state=42)
results = []
for train_indices, test_indices in kf:
train_x = [tr_X[ii] for ii in train_indices]
train_y = [tr_y[ii] for ii in train_indices]
test_x = [tr_X[ii] for ii in test_indices]
test_y = [tr_y[ii] for ii in test_indices]
aud_audio.extract(feature,
cfg.wav_eva_fd,
cfg.eva_fd + '/' + feature,
'parameters.yaml',
dataset=dataset)
############# LOAD DATA ###########################
tr_X, tr_y = seth.get_train_data()
dimx = tr_X.shape[-2]
dimy = tr_X.shape[-1]
tr_X = aud_utils.mat_3d_to_nd(model, tr_X)
miz = aud_model.Functional_Model(input_neurons=input_neurons,
dropout=dropout,
num_classes=num_classes,
model=model,
dimx=dimx,
nb_filter=nb_filter,
pool_size=pool_size,
dimy=dimy,
loss=loss,
optimizer=optimizer)
if prep == 'dev':
kf = KFold(len(tr_X), folds, shuffle=True, random_state=42)
results = []
for train_indices, test_indices in kf:
train_x = [tr_X[ii] for ii in train_indices]
train_y = [tr_y[ii] for ii in train_indices]
test_x = [tr_X[ii] for ii in test_indices]
test_y = [tr_y[ii] for ii in test_indices]
train_y = to_categorical(train_y, num_classes=len(cfg.labels))
test_y = to_categorical(test_y, num_classes=len(cfg.labels))
tr_X = aud_utils.mat_3d_to_nd(model, tr_X)
print(tr_X.shape)
dimx = tr_X.shape[-2]
dimy = tr_X.shape[-1]
if prep == 'dev':
print "Number of folds", folds
cross_validation = True
else:
cross_validation = False
## In case of Functional CRNN
miz = aud_model.Functional_Model(model=model,
dimx=dimx,
dimy=dimy,
num_classes=num_classes,
act1=act1,
act2=act2,
act3=act3)
np.random.seed(68)
if cross_validation:
kf = KFold(len(tr_X), folds, shuffle=True, random_state=42)
results = []
for train_indices, test_indices in kf:
train_x = [tr_X[ii] for ii in train_indices]
train_y = [tr_y[ii] for ii in train_indices]
test_x = [tr_X[ii] for ii in test_indices]
test_y = [tr_y[ii] for ii in test_indices]
train_x = np.array(train_x)
cfg.wav_eva_fd,
cfg.eva_fd + '/' + feature,
'parameters.yaml',
dataset=dataset)
############# LOAD DATA ###########################
tr_X, tr_y = seth.get_train_data()
dimx = tr_X.shape[-2]
dimy = tr_X.shape[-1]
tr_X = aud_utils.mat_3d_to_nd(model, tr_X)
miz = aud_model.Functional_Model(input_neurons=input_neurons,
dropout=dropout,
num_classes=num_classes,
model=model,
dimx=dimx,
nb_filter=nb_filter,
act1=act1,
act2=act2,
act3=act3,
filter_length=filter_length,
dimy=dimy)
if prep == 'dev':
kf = KFold(len(tr_X), folds, shuffle=True, random_state=42)
results = []
for train_indices, test_indices in kf:
train_x = [tr_X[ii] for ii in train_indices]
train_y = [tr_y[ii] for ii in train_indices]
test_x = [tr_X[ii] for ii in test_indices]
test_y = [tr_y[ii] for ii in test_indices]
train_y = to_categorical(train_y, num_classes=len(cfg.labels))
print(tr_X.shape)
print(tr_y.shape)
tr_X=aud_utils.mat_3d_to_nd(model,tr_X)
print(tr_X.shape)
dimx=tr_X.shape[-2]
dimy=tr_X.shape[-1]
if prep=='dev':
print "Number of folds",folds
cross_validation=True
else:
cross_validation=False
nb_filter = [50 , 100]
miz=aud_model.Functional_Model(model=model,dimx=dimx,dimy=dimy,num_classes=num_classes,act1=act1,act2=act2,act3=act3,nb_filter = nb_filter,dropout=dropout1)
np.random.seed(68)
if cross_validation:
kf = KFold(len(tr_X),folds,shuffle=True,random_state=42)
results=[]
for train_indices, test_indices in kf:
train_x = [tr_X[ii] for ii in train_indices]
train_y = [tr_y[ii] for ii in train_indices]
test_x = [tr_X[ii] for ii in test_indices]
test_y = [tr_y[ii] for ii in test_indices]
#train_y = to_categorical(train_y,num_classes=len(labels))
#test_y = to_categorical(test_y,num_classes=len(labels))
train_x=np.array(train_x)
train_y=np.array(train_y)