def ToInputShape(data,net_name,norm=True,test_flag = False):
data = data.astype(np.float32)
batchsize=data.shape[0]
if net_name=='LSTM':
result =[]
for i in range(0,batchsize):
randomdata=random_transform_1d(data[i],finesize = 2700,test_flag=test_flag)
result.append(dsp.getfeature(randomdata))
result = np.array(result).reshape(batchsize,2700*5)
elif net_name=='CNN' or net_name=='resnet18_1d':
result =[]
for i in range(0,batchsize):
randomdata=random_transform_1d(data[i],finesize = 2700,test_flag=test_flag)
# result.append(dsp.getfeature(randomdata,ch_num = 6))
result.append(randomdata)
result = np.array(result)
if norm:
result = Normalize(result,maxmin = 1000,avg=0,sigma=1000)
result = result.reshape(batchsize,1,2700)
elif net_name in ['dfcnn','resnet18','densenet121','densenet201','resnet101','resnet50']:
result =[]
for i in range(0,batchsize):
spectrum = dsp.signal2spectrum(data[i])
spectrum = random_transform_2d(spectrum,(224,122),test_flag=test_flag)
result.append(spectrum)
result = np.array(result)
if norm:
#sleep_def : std,mean,median = 0.4157 0.3688 0.2473
#challge 2018 : std,mean,median,max= 0.2972 0.3008 0.2006 2.0830
result=Normalize(result,2,0.3,1)
result = result.reshape(batchsize,1,224,122)
# print(result.shape)
return result
def ToInputShape(data, opt, test_flag=False):
data = data.astype(np.float32)
batchsize = data.shape[0]
if data.ndim == 2:
loadsize = data.shape[1]
elif data.ndim == 3:
loadsize = data.shape[2]
_finesize = int(loadsize * 0.9)
if opt.model_name == 'lstm':
result = []
for i in range(0, batchsize):
randomdata = random_transform_1d(data[i],
finesize=_finesize,
test_flag=test_flag)
result.append(dsp.getfeature(randomdata))
result = np.array(result).reshape(batchsize, _finesize * 5)
elif opt.model_name in [
'cnn_1d', 'resnet18_1d', 'resnet34_1d', 'multi_scale_resnet_1d',
'micro_multi_scale_resnet_1d'
]:
result = []
for i in range(0, batchsize):
randomdata = random_transform_1d(data[i],
finesize=_finesize,
test_flag=test_flag)
result.append(randomdata)
result = np.array(result)
result = result.reshape(batchsize, opt.input_nc, _finesize)
elif opt.model_name in [
'squeezenet', 'multi_scale_resnet', 'dfcnn', 'resnet18',
'densenet121', 'densenet201', 'resnet101', 'resnet50'
]:
result = []
data = (data - 0.5) * 2
for i in range(0, batchsize):
spectrum = dsp.signal2spectrum(data[i])
spectrum = random_transform_2d(spectrum, (224, 122),
test_flag=test_flag)
result.append(spectrum)
result = np.array(result)
#datasets th_95 avg mid
# sleep_edfx 0.0458 0.0128 0.0053
# CC2018 0.0507 0.0161 0.00828
result = Normalize(result, maxmin=0.5, avg=0.0150, sigma=0.0500)
result = result.reshape(batchsize, 1, 224, 122)
return result