def sample(self, loaders):
args = self.args
nets_ema = self.nets_ema
os.makedirs(args.result_dir, exist_ok=True)
self._load_checkpoint(args.resume_iter)
src = next(InputFetcher(loaders.src, None, args.latent_dim, 'test'))
ref = next(InputFetcher(loaders.ref, None, args.latent_dim, 'test'))
fname = ospj(args.result_dir, 'reference.jpg')
print('Working on {}...'.format(fname))
utils.translate_using_reference(nets_ema, args, src.x, ref.x, ref.y,
fname)
fname = ospj(args.result_dir, 'video_ref.mp4')
print('Working on {}...'.format(fname))
utils.video_ref(nets_ema, args, src.x, ref.x, ref.y, fname)
N = src.x.size(0)
y_trg_list = [
torch.tensor(y).repeat(N).to(device)
for y in range(min(args.num_domains, 5))
]
z_trg_list = torch.randn(args.num_outs_per_domain, 1,
args.latent_dim).repeat(1, N, 1).to(device)
for psi in [0.5, 0.7, 1.0]:
filename = ospj(args.sample_dir,
'%06d_latent_psi_%.1f.jpg' % (step, psi))
translate_using_latent(nets, args, src.x, y_trg_list, z_trg_list,
psi, fname)
fname = ospj(args.result_dir, 'latent.jpg')
print('Working on {}...'.format(fname))
utils.video_ref(nets_ema, args, src.x, ref.x, ref.y, fname)
def sample(self, loaders):
args = self.args
nets_ema = self.nets_ema
os.makedirs(args.result_dir, exist_ok=True)
self._load_checkpoint(args.resume_iter)
src = next(InputFetcher(loaders.src, None, args.latent_dim, 'test'))
ref = next(InputFetcher(loaders.ref, None, args.latent_dim, 'test'))
fname = ospj(args.result_dir, 'reference.jpg')
print('Working on {}...'.format(fname))
utils.translate_using_reference(nets_ema, args, src.x, ref.x, ref.y, fname)
def custom(self, loaders):
args = self.args
nets_ema = self.nets_ema
self._load_checkpoint(100000)
src = next(InputFetcher(loaders.src, None, args.latent_dim, 'test'))
ref = next(InputFetcher(loaders.ref, None, args.latent_dim, 'test'))
fname = args.custom_out_img
print('Working on {}...'.format(fname))
utils.translate_using_reference(nets_ema, args, src.x, ref.x, ref.y,
fname)
def sample(self, loaders):
args = self.args
nets_ema = self.nets_ema
os.makedirs(args.result_dir, exist_ok=True)
self._load_checkpoint(args.resume_iter)
fetch_src = InputFetcher(loaders.src, None, args.latent_dim, 'test')
fetch_ref = InputFetcher(loaders.ref, None, args.latent_dim, 'test')
for i in range(10000):
src = next(fetch_src)
ref = next(fetch_ref)
fname = ospj(
args.result_dir,
str(i).zfill(5) + "from" + str(src.y.item()) + "to" +
str(ref.y.item()) + '.jpg')
print('Working on {}...'.format(fname))
utils.translate_using_reference(nets_ema, args, src.x, src.y,
ref.x, ref.y, fname)
def sample(self,src,ref):
args = self.args
nets_ema = self.nets_ema
os.makedirs(args.result_dir, exist_ok=True)
self._load_checkpoint(args.resume_iter)
src = self.convert_to_tensor(src)
ref = self.convert_to_tensor(ref)
fname = ospj(args.result_dir, 'reference.jpg')
print('Working on {}...'.format(fname))
images_all = utils.translate_using_reference(nets_ema, args, src, ref, torch.tensor([1]), fname)
image = images_all[0,...]
image = image.permute(1,2,0)
image = image.cpu().numpy()*255
image = cv2.cvtColor(image,cv2.COLOR_RGB2BGR)
return image
