def wav_loader(path):
#print(path)
npys = np.zeros((300, 64))
temp = audio_processing.mk_MFB(path)
l = temp.shape[0]
if l <= 300:
npys[:l, :] = temp
npys[l:, :] = temp[:300 - l, :]
else:
randint = np.random.randint(l - 300)
npys = temp[randint:randint + 300, :]
mu = np.average(npys)
sigma = np.std(npys)
npys = (npys - mu) / max(sigma, 0.001)
#npys = (npys - np.min(npys))/(np.max(npys)-np.min(npys))
return npys
def forward(self, anchor, positive, negative):
d_p = self.pdist.forward(anchor, positive)
d_n = self.pdist.forward(anchor, negative)
dist_hinge = torch.clamp(self.margin + d_p - d_n, min=0.0)
loss = torch.mean(dist_hinge)
return loss
kwargs = {'num_workers': 0, 'pin_memory': True} if args.cuda else {}
l2_dist = PairwiseDistance(2)
voxceleb = read_voxceleb_structure(args.dataroot)
if args.makemfb:
for datum in voxceleb.iterrows():
mk_MFB((args.dataroot +'/voxceleb1_wav/' + datum[1]['filename']+'.wav'))
if args.mfb:
transform = transforms.Compose([
truncatedinputfromMFB(),
totensor()
])
file_loader = read_MFB
else:
transform = transforms.Compose([
truncatedinput(),
toMFB(),
totensor(),
#tonormal()
])
file_loader = read_audio