def get_current_visuals(self):
real_A = util.tensor2im(self.real_A.data)
fake_B = util.tensor2im(self.fake_B.data)
real_B = util.tensor2im(self.real_B.data)
real_AA = util.tensor2im(self.real_AA.data)
fake_BB = util.tensor2im(self.fake_BB.data)
real_BB = util.tensor2im(self.real_BB.data)
if self.output_str_indices:
output_str_indices = torch.LongTensor(self.output_str_indices)
output_str_data = self.fake_BB.data[:, output_str_indices, :, :]
real_str_data = self.real_BB.data[:, output_str_indices, :, :]
img_output_str = util.tensor2im(output_str_data)
img_real_str = util.tensor2im(real_str_data)
list = [('Ereal_A', real_A), ('Ereal_B', real_B),
('Efake_B', fake_B), ('real_A', real_AA),
('real_B', real_BB), ('fake_B', fake_BB),
('fake_STR', img_output_str), ('real_STR', img_real_str)]
else:
list = [('Ereal_A', real_A), ('Ereal_B', real_B),
('Efake_B', fake_B), ('real_A', real_AA),
('real_B', real_BB), ('fake_B', fake_BB)]
if self.opt.str_input:
list = [('real_A', real_AA), ('real_B', real_BB),
('fake_B', fake_BB)]
return OrderedDict(list)
def get_current_visuals(self):
real_A1 = self.real_A1.data.clone()
g, c, m, n = real_A1.size()
fake_B = self.fake_B1.data.clone()
real_B = self.real_B1.data.clone()
# if self.opt.isTrain:
# real_A_all = real_A1
# fake_B_all = fake_B
# else:
# real_A_all = self.Tensor(real_B.size(0), real_B.size(1),
# real_A1.size(2),
# real_A1.size(2) * real_A1.size(
# 0))
# fake_B_all = self.Tensor(real_B.size(0), real_B.size(1),
# real_A1.size(2),
# fake_B.size(2) * fake_B.size(0))
# for b in range(g):
# real_A_all[:, :, :, self.out_id[b] * m:m * (
# self.out_id[b] + 1)] = real_A1[b, :, :, :]
# fake_B_all[:, :, :,
# self.out_id[b] * m:m * (self.out_id[b] + 1)] = fake_B[
# b, :,
# :, :]
real_A_all = self.Tensor(real_B.size(0), real_B.size(1),
real_A1.size(2),
real_A1.size(2) * real_A1.size(0))
fake_B_all = self.Tensor(real_B.size(0), real_B.size(1),
real_A1.size(2),
fake_B.size(2) * fake_B.size(0))
for b in range(g):
real_A_all[:, :, :, b * m:m * (b + 1)] = real_A1[b, :, :, :]
fake_B_all[:, :, :, b * m:m * (b + 1)] = fake_B[b, :, :, :]
real_A = util.tensor2im(real_A_all)
fake_B = util.tensor2im(fake_B_all)
real_B = util.tensor2im(self.real_B1.data)
return OrderedDict([('real_A', real_A), ('fake_B', fake_B),
('real_B', real_B)])
def forward(self):
self.real_A = Variable(self.input_A)
# print('REAL_A size is')
# print(self.real_A.size())
if self.opt.conv3d:
self.real_A_indep = self.netG_3d.forward(self.real_A.unsqueeze(2))
# print('REAL_A_INDEP size is')
# print(self.real_A_indep.size())
self.fake_B = self.netG.forward(self.real_A_indep.squeeze(2))
else:
self.fake_B = self.netG.forward(self.real_A)
# print('FAKE_B size is')
# print(self.fake_B.size())
b, c, m, n = self.fake_B.size()
self.real_AA = self.real_A.narrow(3, m, m)
self.fake_BB = self.fake_B.narrow(3, m, m)
self.real_B = Variable(self.input_B)
self.real_BB = self.real_B.narrow(3, m, m)
# print('REAL_B size is')
# print(self.real_B.size())
real_B = util.tensor2im(self.real_B.data)
real_A = util.tensor2im(self.real_A.data)
def parse(self):
if not self.initialized:
self.initialize()
self.opt = self.parser.parse_args()
self.opt.isTrain = self.isTrain # train or test
self.opt.project_root = self.project_root
self.opt.computer_name = self.computer_name
if self.opt.gpu_ids.find(' ') == -1:
gpu_id_list = self.opt.gpu_ids.split(',')
self.opt.gpu_ids = []
if len(gpu_id_list) >= 1:
for gpu_id in gpu_id_list:
gpu_id = int(gpu_id)
if gpu_id >= 0:
self.opt.gpu_ids.append(gpu_id)
else:
self.opt.gpu_ids = []
args = vars(self.opt)
if self.opt.dataset == 'SandunLK10k64':
self.opt.charset = 'ENfull'
elif self.opt.dataset == 'Capitals64':
self.opt.charset = 'ENcap'
else:
print('Unknown dataset!!!')
self.opt.charset = 'ENcap'
if self.opt.charset == 'ENcap':
self.opt.input_nc = 26
self.opt.output_nc = 26
self.opt.grps = 26
if self.opt.blanks == -1:
self.opt.blanks = 0.7
elif self.opt.charset == 'ENfull':
self.opt.input_nc = 60
self.opt.output_nc = 60
self.opt.grps = 60
if self.opt.blanks == -1:
self.opt.blanks = 0.85
# save to the disk
expr_dir = os.path.join(".", self.opt.checkpoints_dir,
self.opt.experiment_dir)
XIutil.mkdirs(expr_dir)
file_name = os.path.join(expr_dir, 'opt.txt')
if self.opt.isTrain:
self.opt.str_input_indices = []
else:
self.opt.str_input_indices = str2index(
self.opt.str_input, self.opt.charset)
self.opt.str_output_indices = str2index(self.opt.str_input, self.opt.charset)
if self.opt.zip_results == -1:
if self.opt.isTrain:
self.opt.zip_results = 0
else:
self.opt.zip_results = 1
print('------------ Options -------------')
for k, v in sorted(args.items()):
print('%s: %s' % (str(k), str(v)))
print('-------------- End ----------------')
with open(file_name, 'wt') as opt_file:
opt_file.write('------------ Options -------------\n')
for k, v in sorted(args.items()):
opt_file.write('%s: %s\n' % (str(k), str(v)))
opt_file.write('-------------- End ----------------\n')
opt_file.close()
if self.opt.use_auxiliary:
auxiliary_expr_dir = os.path.join(self.opt.auxiliary_root,
self.opt.project_relative,
self.opt.checkpoints_dir,
self.opt.experiment_dir)
print(auxiliary_expr_dir)
XIutil.mkdir(auxiliary_expr_dir)
copy2(file_name, auxiliary_expr_dir)
return self.opt