def train_tryon(opt, train_loader, model):
model.cuda()
model.train()
for step in range(opt.keep_step + opt.decay_step):
iter_start_time = time.time()
inputs = train_loader.next_batch()
print(step, '++++++')
print("-------------------------------------")
pairs = inputs["pair"]
# c_name = inputs["c_name"]
model.set_input(inputs)
model.optimize_parameters()
results = model.current_results()
print('step: %8d, G_loss: %4f, c_loss: %4f' %
(step + 1, results['G_loss'].item(),
results['content_loss'].item()),
flush=True)
isExists = os.path.exists(os.path.join(opt.result, opt.name))
if not isExists:
os.makedirs(os.path.join(opt.result, opt.name))
save_images(results['gen_B'], pairs,
os.path.join(opt.result, opt.name))
# # print(name1)
if (step + 1) % 100 == 0:
save_checkpoint(
model,
os.path.join(opt.checkpoint_dir, opt.name,
'step_%06d.pth' % (step + 1)))
def train_gmm(opt, train_loader, model, board):
if opt.cuda:
model.cuda()
model.train()
# criterion
criterionL1 = nn.L1Loss()
# optimizer
optimizer = torch.optim.Adam(model.parameters(), lr=opt.lr, betas=(0.5, 0.999))
scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer, lr_lambda = lambda step: 1.0 -
max(0, step - opt.keep_step) / float(opt.decay_step + 1))
for step in range(opt.keep_step + opt.decay_step):
iter_start_time = time.time()
inputs = train_loader.next_batch()
if opt.cuda:
im = inputs['image'].cuda()
im_pose = inputs['pose_image'].cuda()
im_h = inputs['head'].cuda()
shape = inputs['shape'].cuda()
agnostic = inputs['agnostic'].cuda()
c = inputs['cloth'].cuda()
cm = inputs['cloth_mask'].cuda()
im_c = inputs['parse_cloth'].cuda()
im_g = inputs['grid_image'].cuda()
else:
im = inputs['image']
im_pose = inputs['pose_image']
im_h = inputs['head']
shape = inputs['shape']
agnostic = inputs['agnostic']
c = inputs['cloth']
cm = inputs['cloth_mask']
im_c = inputs['parse_cloth']
im_g = inputs['grid_image']
grid, theta = model(agnostic, c)
warped_cloth = F.grid_sample(c, grid, padding_mode='border')
warped_mask = F.grid_sample(cm, grid, padding_mode='zeros')
warped_grid = F.grid_sample(im_g, grid, padding_mode='zeros')
visuals = [ [im_h, shape, im_pose],
[c, warped_cloth, im_c],
[warped_grid, (warped_cloth+im)*0.5, im]]
loss = criterionL1(warped_cloth, im_c) #$$
optimizer.zero_grad()
loss.backward()
optimizer.step()
if (step+1) % opt.display_count == 0:
board_add_images(board, 'combine', visuals, step+1)
board.add_scalar('metric', loss.item(), step+1)
t = time.time() - iter_start_time
print('step: %8d, time: %.3f, loss: %4f' % (step+1, t, loss.item()), flush=True)
if (step+1) % opt.save_count == 0:
save_checkpoint(model, os.path.join(opt.checkpoint_dir, opt.name, 'step_%06d.pth' % (step+1)))
def main():
opt = get_opt()
# opt.cuda = False
# opt.batch_size = 1
# opt.name = "test"
print(opt)
print("Start to train stage: %s, named: %s!" % (opt.stage, opt.name))
# create dataset
train_dataset = CPDataset(opt)
# create dataloader
train_loader = CPDataLoader(opt, train_dataset)
# visualization
if not os.path.exists(opt.tensorboard_dir):
os.makedirs(opt.tensorboard_dir)
board = SummaryWriter(log_dir=os.path.join(opt.tensorboard_dir, opt.name))
# create model & train & save the final checkpoint
G = WUTON(opt)
D = Discriminator()
if not opt.checkpoint == '' and os.path.exists(opt.checkpoint): # TODO
load_checkpoint(G, opt.checkpoint)
train(opt, train_loader, G, D, board)
# train2(opt, train_loader, G, board)
save_checkpoint(
G, os.path.join(opt.checkpoint_dir, opt.name, 'wuton_final.pth'))
print('Finished training %s, named: %s!' % (opt.stage, opt.name))
def main():
opt = get_opt()
print(opt)
print("Start to train stage: %s, named: %s!" % (opt.stage, opt.name))
# create dataset
train_dataset = CPDataset(opt)
# create dataloader
train_loader = CPDataLoader(opt, train_dataset)
# visualization
if not os.path.exists(opt.tensorboard_dir):
os.makedirs(opt.tensorboard_dir)
board = SummaryWriter(log_dir = os.path.join(opt.tensorboard_dir, opt.name))
# create model & train & save the final checkpoint
if opt.stage == 'GMM':
model = GMM(opt)
if not opt.checkpoint =='' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
train_gmm(opt, train_loader, model, board)
save_checkpoint(model, os.path.join(opt.checkpoint_dir, opt.name, 'gmm_final.pth'))
elif opt.stage == 'TOM':
model = UnetGenerator(25, 4, 6, ngf=64, norm_layer=nn.InstanceNorm2d)
if not opt.checkpoint =='' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
train_tom(opt, train_loader, model, board)
save_checkpoint(model, os.path.join(opt.checkpoint_dir, opt.name, 'tom_final.pth'))
else:
raise NotImplementedError('Model [%s] is not implemented' % opt.stage)
print('Finished training %s, nameed: %s!' % (opt.stage, opt.name))
def train_tom(opt, train_loader, model, board):
model.cuda()
model.train()
# criterion
criterionL1 = nn.L1Loss()
criterionVGG = VGGLoss()
criterionMask = nn.L1Loss()
# optimizer
optimizer = torch.optim.Adam(model.parameters(), lr=opt.lr, betas=(0.5, 0.999))
scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer, lr_lambda = lambda step: 1.0 -
max(0, step - opt.keep_step) / float(opt.decay_step + 1))
for step in range(opt.keep_step + opt.decay_step):
iter_start_time = time.time()
inputs = train_loader.next_batch()
im = inputs['image'].cuda()
im_pose = inputs['pose_image']
im_h = inputs['head']
shape = inputs['shape']
agnostic = inputs['agnostic'].cuda()
c = inputs['cloth'].cuda()
cm = inputs['cloth_mask'].cuda()
outputs = model(torch.cat([agnostic, c],1))
p_rendered, m_composite = torch.split(outputs, 3,1)
p_rendered = F.tanh(p_rendered)
m_composite = F.sigmoid(m_composite)
p_tryon = c * m_composite+ p_rendered * (1 - m_composite)
visuals = [ [im_h, shape, im_pose],
[c, cm*2-1, m_composite*2-1],
[p_rendered, p_tryon, im]]
loss_l1 = criterionL1(p_tryon, im)
loss_vgg = criterionVGG(p_tryon, im)
loss_mask = criterionMask(m_composite, cm)
loss = loss_l1 + loss_vgg + loss_mask
optimizer.zero_grad()
loss.backward()
optimizer.step()
if (step+1) % opt.display_count == 0:
board_add_images(board, 'combine', visuals, step+1)
board.add_scalar('metric', loss.item(), step+1)
board.add_scalar('L1', loss_l1.item(), step+1)
board.add_scalar('VGG', loss_vgg.item(), step+1)
board.add_scalar('MaskL1', loss_mask.item(), step+1)
#board.add_graph(model, torch.cat([agnostic, c],1))
t = time.time() - iter_start_time
print('step: %8d, time: %.3f, loss: %.4f, l1: %.4f, vgg: %.4f, mask: %.4f'
% (step+1, t, loss.item(), loss_l1.item(),
loss_vgg.item(), loss_mask.item()), flush=True)
if (step+1) % opt.save_count == 0:
save_checkpoint(model, os.path.join(opt.checkpoint_dir, opt.name, 'step_%06d.pth' % (step+1)))
def main():
opt = get_opt()
print(opt)
print('Loading dataset')
dataset_train = TOMDataset(opt, mode='train', data_list='train_pairs.txt')
dataloader_train = DataLoader(dataset_train,
batch_size=opt.batch_size,
num_workers=opt.n_worker,
shuffle=True)
dataset_val = TOMDataset(opt,
mode='val',
data_list='val_pairs.txt',
train=False)
dataloader_val = DataLoader(dataset_val,
batch_size=opt.batch_size,
num_workers=opt.n_worker,
shuffle=True)
save_dir = os.path.join(opt.out_dir, opt.name)
log_dir = os.path.join(opt.out_dir, 'log')
dirs = [opt.out_dir, save_dir, os.path.join(save_dir, 'train'), log_dir]
for d in dirs:
mkdir(d)
log_name = os.path.join(log_dir, opt.name + '.csv')
with open(log_name, 'w') as f:
f.write('epoch,train_loss,val_loss\n')
print('Building TOM model')
gen = UnetGenerator(25, 4, 6, ngf=64, norm_layer=nn.InstanceNorm2d)
dis = NLayerDiscriminator(28,
ndf=64,
n_layers=6,
norm_layer=nn.InstanceNorm2d,
use_sigmoid=True)
gen.cuda()
dis.cuda()
n_step = int(opt.n_epoch * len(dataset_train) / opt.batch_size)
trainer = TOMTrainer(gen, dis, dataloader_train, dataloader_val,
opt.gpu_id, opt.log_freq, save_dir, n_step)
print('Start training TOM')
for epoch in tqdm(range(opt.n_epoch)):
print('Epoch: {}'.format(epoch))
loss = trainer.train(epoch)
print('Train loss: {:.3f}'.format(loss))
with open(log_name, 'a') as f:
f.write('{},{:.3f},'.format(epoch, loss))
save_checkpoint(
gen, os.path.join(save_dir, 'gen_epoch_{:02}.pth'.format(epoch)))
save_checkpoint(
dis, os.path.join(save_dir, 'dis_epoch_{:02}.pth'.format(epoch)))
loss = trainer.val(epoch)
print('Validation loss: {:.3f}'.format(loss))
with open(log_name, 'a') as f:
f.write('{:.3f}\n'.format(loss))
print('Finish training TOM')
def train_identity_embedding(opt, train_loader, model, board):
model.cuda()
model.train()
# criterion
mse_criterion = torch.nn.MSELoss()
triplet_criterion = torch.nn.TripletMarginLoss(margin=0.3)
# optimizer
optimizer = torch.optim.Adam(model.parameters(), lr=opt.lr)
pbar = range(opt.keep_step + opt.decay_step)
if single_gpu_flag(opt):
pbar = tqdm(pbar)
for step in pbar:
inputs_1, inputs_2 = train_loader.next_batch()
img_1 = inputs_1['cloth'].cuda()
img_ou_1 = inputs_1['image'].cuda()
img_2 = inputs_2['cloth'].cuda()
img_ou_2 = inputs_2['image'].cuda()
pred_prod_embedding_1, pred_outfit_embedding_1 = model(img_1, img_ou_1)
pred_prod_embedding_2, pred_outfit_embedding_2 = model(img_2, img_ou_2)
# msee loss
mean_squared_loss = (
mse_criterion(pred_outfit_embedding_1, pred_prod_embedding_1) +
mse_criterion(pred_outfit_embedding_2, pred_prod_embedding_2)) / 2
# triplet loss
triplet_loss = triplet_criterion(pred_outfit_embedding_1, pred_prod_embedding_1, pred_outfit_embedding_2) + \
triplet_criterion(pred_outfit_embedding_2, pred_prod_embedding_2, pred_outfit_embedding_1) + \
triplet_criterion(pred_outfit_embedding_1, pred_prod_embedding_1, pred_prod_embedding_2) + \
triplet_criterion(pred_outfit_embedding_2, pred_prod_embedding_2, pred_prod_embedding_1)
loss = mean_squared_loss + triplet_loss
optimizer.zero_grad()
loss.backward()
optimizer.step()
if single_gpu_flag(opt):
board.add_scalar('metric', loss.item(), step + 1)
board.add_scalar('MSE', mean_squared_loss.item(), step + 1)
board.add_scalar('trip', triplet_loss.item(), step + 1)
pbar.set_description(
'step: %8d, loss: %.4f, mse: %.4f, trip: %.4f' %
(step + 1, loss.item(), mean_squared_loss.item(),
triplet_loss.item()))
if (step + 1) % opt.save_count == 0 and single_gpu_flag(opt):
save_checkpoint(
model,
os.path.join(opt.checkpoint_dir, opt.name,
'step_%06d.pth' % (step + 1)))
def train_gmm(opt, train_loader, model, board):
model.cuda()
model.train()
# criterion
criterionL1 = nn.L1Loss()
# optimizer
optimizer = torch.optim.Adam(model.parameters(),
lr=opt.lr,
betas=(0.5, 0.999))
scheduler = torch.optim.lr_scheduler.LambdaLR(
optimizer,
lr_lambda=lambda step: 1.0 - max(0, step - opt.keep_step) / float(
opt.decay_step + 1))
for step in range(opt.keep_step + opt.decay_step):
iter_start_time = time.time()
inputs = train_loader.next_batch()
c_image = inputs['c_image'].cuda()
t_pose = inputs['t_pose'].cuda()
c_head = inputs['c_head'].cuda()
Pre_target_mask = inputs['Pre_target_mask'].cuda()
cloth = inputs['cloth'].cuda()
cloth_mask = inputs['cloth_mask'].cuda()
t_upper_mask = inputs['t_upper_mask'].cuda()
t_upper_cloth = inputs['t_upper_cloth'].cuda()
im_g = inputs['grid_image'].cuda()
# agnostic = inputs['agnostic'].cuda()
agnostic = torch.cat([Pre_target_mask, t_pose], 1)
grid, theta = model(agnostic, cloth)
warped_cloth = F.grid_sample(cloth, grid, padding_mode='border')
warped_mask = F.grid_sample(cloth_mask, grid, padding_mode='zeros')
warped_grid = F.grid_sample(im_g, grid, padding_mode='zeros')
visuals = [[c_head, Pre_target_mask, t_pose],
[cloth_mask, warped_cloth, t_upper_mask],
[warped_grid, (warped_cloth + c_image) * 0.5, c_image]]
loss = criterionL1(warped_cloth, t_upper_cloth)
optimizer.zero_grad()
loss.backward()
optimizer.step()
if (step + 1) % opt.display_count == 0:
t = time.time() - iter_start_time
print('step: %8d, time: %.3f, loss: %4f' %
(step + 1, t, loss.item()),
flush=True)
if (step + 1) % opt.save_count == 0:
save_checkpoint(
model,
os.path.join(opt.checkpoint_dir, opt.name,
'step_%06d.pth' % (step + 1)))
def main():
opt = get_opt()
print(opt)
print("Start to train stage: %s, named: %s!" % (opt.stage, opt.name))
# create dataset
train_dataset = CPDataset(opt)
# create dataloader
train_loader = CPDataLoader(opt, train_dataset)
# visualization
if not os.path.exists(opt.tensorboard_dir):
os.makedirs(opt.tensorboard_dir)
board = SummaryWriter(log_dir=os.path.join(opt.tensorboard_dir, opt.name))
# create model & train & save the final checkpoint
if opt.stage == 'GMM':
model = GMM(opt, 1)
if not opt.checkpoint == '' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
train_gmm(opt, train_loader, model, board)
save_checkpoint(
model, os.path.join(opt.checkpoint_dir, opt.name, 'gmm_final.pth'))
elif opt.stage == 'HPM':
model = UnetGenerator(25,
16,
6,
ngf=64,
norm_layer=nn.InstanceNorm2d,
clsf=True)
d_g = Discriminator_G(opt, 16)
if not opt.checkpoint == '' and os.path.exists(opt.checkpoint):
if not os.path.isdir(opt.checkpoint):
raise NotImplementedError(
'checkpoint should be dir, not file: %s' % opt.checkpoint)
load_checkpoints(model, d_g, os.path.join(opt.checkpoint,
"%s.pth"))
train_hpm(opt, train_loader, model, d_g, board)
save_checkpoints(
model, d_g,
os.path.join(opt.checkpoint_dir,
opt.stage + '_' + opt.name + "_final", '%s.pth'))
elif opt.stage == 'TOM':
model = UnetGenerator(25, 3, 6, ngf=64, norm_layer=nn.InstanceNorm2d)
if not opt.checkpoint == '' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
train_tom(opt, train_loader, model, board)
save_checkpoint(
model, os.path.join(opt.checkpoint_dir, opt.name, 'tom_final.pth'))
else:
raise NotImplementedError('Model [%s] is not implemented' % opt.stage)
print('Finished training %s, nameed: %s!' % (opt.stage, opt.name))
def train_mask_gen(opt, train_loader, model):
model.cuda()
model.train()
# criterion
criterionL1 = nn.L1Loss()
# optimizer
optimizer = torch.optim.Adam(model.parameters(),
lr=opt.lr,
betas=(0.5, 0.999))
scheduler = torch.optim.lr_scheduler.LambdaLR(
optimizer,
lr_lambda=lambda step: 1.0 - max(0, step - opt.keep_step) / float(
opt.decay_step + 1))
for step in range(opt.keep_step + opt.decay_step):
iter_start_time = time.time()
inputs = train_loader.next_batch()
c = inputs['cloth'].cuda()
mesh = inputs['mesh'].cuda()
pose_map = inputs['pose_map'].cuda()
person_parse = inputs['person_parse'].cuda()
outputs = model(torch.cat([mesh, pose_map, c], 1))
#p_rendered, m_composite = torch.split(outputs, 3,1)
#p_rendered = F.tanh(p_rendered)
m_composite = F.sigmoid(outputs)
#p_tryon = c * m_composite + p_rendered * (1 - m_composite)
loss_l1 = criterionL1(m_composite, person_parse)
loss = loss_l1
optimizer.zero_grad()
loss.backward()
optimizer.step()
if (step + 1) % opt.display_count == 0:
t = time.time() - iter_start_time
print('step: %8d, time: %.3f, loss: %.4f' %
(step + 1, t, loss.item()),
flush=True)
if (step + 1) % opt.save_count == 0:
save_checkpoint(
model,
os.path.join('./mask_gen/', 'model_',
'step_%06d.pth' % (step + 1)))
print('Finished in ' + str(time.time() - start_time))
def train_gmm(opt, train_loader, model, board):
model.cuda()
model.train()
# criterion
criterionL1 = nn.L1Loss()
# optimizer
optimizer = torch.optim.Adam(model.parameters(),
lr=opt.lr,
betas=(0.5, 0.999))
scheduler = torch.optim.lr_scheduler.LambdaLR(
optimizer,
lr_lambda=lambda step: 1.0 - max(0, step - opt.keep_step) / float(
opt.decay_step + 1))
#change
loss_sum = 0
for step in range(opt.keep_step + opt.decay_step):
iter_start_time = time.time()
inputs = train_loader.next_batch()
im = inputs['image'].cuda()
c = inputs['cloth'].cuda()
cm = inputs['cloth_mask'].cuda()
wm = inputs['warped_mask'].cuda()
im_c = inputs['parse_cloth'].cuda()
im_g = inputs['grid_image'].cuda()
grid, theta = model(wm, c)
warped_cloth = F.grid_sample(c, grid, padding_mode='border')
warped_mask = F.grid_sample(cm, grid, padding_mode='zeros')
warped_grid = F.grid_sample(im_g, grid, padding_mode='zeros')
loss = criterionL1(warped_mask, wm)
optimizer.zero_grad()
loss.backward()
optimizer.step()
loss_sum += loss.item()
if (step + 1) % opt.display_count == 0:
board.add_scalar('metric', loss.item(), step + 1)
t = time.time() - iter_start_time
print('step: %8d, time: %.3f, loss: %.4f' %
(step + 1, t, loss.item()),
flush=True)
if (step + 1) % opt.save_count == 0:
save_checkpoint(
model,
os.path.join(opt.checkpoint_dir, opt.name,
'step_%06d.pth' % (step + 1)))
def main():
opt = get_opt()
print(opt)
print('Loading dataset')
dataset_train = GMMDataset(opt, mode='train', data_list='train_pairs.txt')
dataloader_train = DataLoader(dataset_train,
batch_size=opt.batch_size,
num_workers=opt.n_worker,
shuffle=True)
dataset_val = GMMDataset(opt,
mode='val',
data_list='val_pairs.txt',
train=False)
dataloader_val = DataLoader(dataset_val,
batch_size=opt.batch_size,
num_workers=opt.n_worker,
shuffle=True)
save_dir = os.path.join(opt.out_dir, opt.name)
log_dir = os.path.join(opt.out_dir, 'log')
dirs = [opt.out_dir, save_dir, os.path.join(save_dir, 'train'), log_dir]
for d in dirs:
mkdir(d)
log_name = os.path.join(log_dir, opt.name + '.csv')
with open(log_name, 'w') as f:
f.write('epoch,train_loss,val_loss\n')
print('Building GMM model')
model = GMM(opt)
model.cuda()
trainer = GMMTrainer(model, dataloader_train, dataloader_val, opt.gpu_id,
opt.log_freq, save_dir)
print('Start training GMM')
for epoch in tqdm(range(opt.n_epoch)):
print('Epoch: {}'.format(epoch))
loss = trainer.train(epoch)
print('Train loss: {:.3f}'.format(loss))
with open(log_name, 'a') as f:
f.write('{},{:.3f},'.format(epoch, loss))
save_checkpoint(
model, os.path.join(save_dir, 'epoch_{:02}.pth'.format(epoch)))
loss = trainer.val(epoch)
print('Validation loss: {:.3f}'.format(loss))
with open(log_name, 'a') as f:
f.write('{:.3f}\n'.format(loss))
print('Finish training GMM')
def main():
opt = get_opt()
print(opt)
print("named: %s!" % (opt.name))
# create dataset
train_dataset = CPDataset(opt)
generator = cyclegan(opt)
print('/////////////////////////////////')
# create dataloader
train_loader = CPDataLoader(opt, train_dataset)
print('/////////////////////////////////')
if not opt.checkpoint == '' and os.path.exists(opt.checkpoint):
load_checkpoint(generator, opt.checkpoint)
print('/////////////////////////////////')
train_tryon(opt, train_loader, generator)
save_checkpoint(generator,
os.path.join(opt.checkpoint_dir, opt.name, 'try_on.pth'))
def main():
opt = get_opt()
print(opt)
print("Start to train stage: %s, named: %s!" % (opt.stage, opt.name))
# create dataset
train_dataset = get_dataset_class(opt.dataset)(opt)
# create dataloader
train_loader = CPDataLoader(opt, train_dataset)
# visualization
board = None
if opt.tensorboard_dir and not os.path.exists(opt.tensorboard_dir):
os.makedirs(opt.tensorboard_dir)
board = SummaryWriter(log_dir=os.path.join(opt.tensorboard_dir, opt.name))
# create model & train & save the final checkpoint
if opt.stage == "GMM":
model = GMM(opt)
model.opt = opt
if not opt.checkpoint == "" and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
if torch.cuda.device_count() > 1:
model = nn.DataParallel(model)
train_gmm(opt, train_loader, model, board)
save_checkpoint(
model, os.path.join(opt.checkpoint_dir, opt.name, "gmm_final.pth"))
elif opt.stage == "TOM":
model = UnetGenerator(25, 4, 6, ngf=64, norm_layer=nn.InstanceNorm2d)
model.opt = opt
if not opt.checkpoint == "" and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
if torch.cuda.device_count() > 1:
model = nn.DataParallel(model)
train_tom(opt, train_loader, model, board)
save_checkpoint(
model, os.path.join(opt.checkpoint_dir, opt.name, "tom_final.pth"))
else:
raise NotImplementedError("Model [%s] is not implemented" % opt.stage)
print("Finished training %s, nameed: %s!" % (opt.stage, opt.name))
def main():
opt = get_opt()
print(opt)
print("Start to train stage: %s, named: %s!" % (opt.stage, opt.name))
# create dataset
train_dataset = CPDataset(opt)
# create dataloader
train_loader = CPDataLoader(opt, train_dataset)
print('//////////////////////')
# visualization
# if not os.path.exists(opt.tensorboard_dir):
# os.makedirs(opt.tensorboard_dir)
# # board = SummaryWriter(log_dir = 'G:/work 2/Codes/cp-vton-master/tensorboard')
if not os.path.exists(opt.tensorboard_dir):
os.makedirs(opt.tensorboard_dir)
board = SummaryWriter(log_dir=os.path.join(opt.tensorboard_dir, opt.name))
# create model & train & save the final checkpoint
if opt.stage == 'PGP':
model = PGP(opt)
if not opt.checkpoint == '' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
train_pgp(opt, train_loader, model, board)
save_checkpoint(
model, os.path.join(opt.checkpoint_dir, opt.name, 'pgp_final.pth'))
elif opt.stage == 'GMM':
model = GMM(opt)
if not opt.checkpoint == '' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
train_gmm(opt, train_loader, model, board)
save_checkpoint(
model, os.path.join(opt.checkpoint_dir, opt.name, 'gmm_final.pth'))
else:
raise NotImplementedError('Model [%s] is not implemented' % opt.stage)
print('Finished training %s, nameed: %s!' % (opt.stage, opt.name))
def train_tom(opt, train_loader, model, board):
# load model
model.cuda()
model.train()
# criterion
criterionL1 = nn.L1Loss()
criterionVGG = VGGLoss()
criterionMask = nn.L1Loss()
# optimizer
optimizer = torch.optim.Adam(model.parameters(), lr=opt.lr, betas=(0.5, 0.999))
scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer, lr_lambda = lambda step: 1.0 -
max(0, step - opt.keep_step) / float(opt.decay_step + 1))
# train log
if not opt.checkpoint == '':
train_log = open(os.path.join(opt.checkpoint_dir, opt.name, 'train_log.txt'), 'a')
else:
os.makedirs(os.path.join(opt.checkpoint_dir, opt.name), exist_ok=True)
train_log = open(os.path.join(opt.checkpoint_dir, opt.name, 'train_log.txt'), 'w')
train_log.write('='*30 + ' Training Option ' + '='*30 + '\n')
train_log.write(str(opt) + '\n\n')
train_log.write('='*30 + ' Network Architecture ' + '='*30 + '\n')
print(str(model) + '\n', file=train_log)
train_log.write('='*30 + ' Training Log ' + '='*30 + '\n')
# train loop
checkpoint_step = 0
if not opt.checkpoint == '':
checkpoint_step += int(opt.checkpoint.split('/')[-1][5:11])
for step in range(checkpoint_step, opt.keep_step + opt.decay_step):
iter_start_time = time.time()
dl_iter = iter(train_loader)
inputs = dl_iter.next()
im = inputs['image'].cuda()
im_pose = inputs['pose_image']
im_h = inputs['head']
shape = inputs['shape']
agnostic = inputs['agnostic'].cuda()
c = inputs['cloth'].cuda()
cm = inputs['cloth_mask'].cuda()
outputs = model(torch.cat([agnostic, c],1))
p_rendered, m_composite = torch.split(outputs, 3,1)
p_rendered = torch.tanh(p_rendered)
m_composite = torch.sigmoid(m_composite)
p_tryon = c * m_composite+ p_rendered * (1 - m_composite)
visuals = [ [im_h, shape, im_pose],
[c, cm*2-1, m_composite*2-1],
[p_rendered, p_tryon, im]]
loss_l1 = criterionL1(p_tryon, im)
loss_vgg = criterionVGG(p_tryon, im)
loss_mask = criterionMask(m_composite, cm)
loss = loss_l1 + loss_vgg + loss_mask
optimizer.zero_grad()
loss.backward()
optimizer.step()
scheduler.step()
if (step+1) % opt.display_count == 0:
board_add_images(board, 'combine', visuals, step+1)
board.add_scalar('metric', loss.item(), step+1)
board.add_scalar('L1', loss_l1.item(), step+1)
board.add_scalar('VGG', loss_vgg.item(), step+1)
board.add_scalar('MaskL1', loss_mask.item(), step+1)
t = time.time() - iter_start_time
print('step: %8d, time: %.3f, loss: %.4f, l1: %.4f, vgg: %.4f, mask: %.4f'
% (step+1, t, loss.item(), loss_l1.item(),
loss_vgg.item(), loss_mask.item()), flush=True)
train_log.write('step: %8d, time: %.3f, loss: %.4f, l1: %.4f, vgg: %.4f, mask: %.4f'
% (step+1, t, loss.item(), loss_l1.item(), loss_vgg.item(), loss_mask.item()) + '\n')
if (step+1) % opt.save_count == 0:
save_checkpoint(model, os.path.join(opt.checkpoint_dir, opt.name, 'step_%06d.pth' % (step+1)))
def train_gmm(opt, train_loader, model, board):
model.cuda()
model.train()
# criterion
criterionL1 = nn.L1Loss()
gicloss = GicLoss(opt)
# optimizer
optimizer = torch.optim.Adam(model.parameters(),
lr=opt.lr,
betas=(0.5, 0.999))
scheduler = torch.optim.lr_scheduler.LambdaLR(
optimizer,
lr_lambda=lambda step: 1.0 - max(0, step - opt.keep_step) / float(
opt.decay_step + 1))
for step in range(opt.keep_step + opt.decay_step):
iter_start_time = time.time()
inputs = train_loader.next_batch()
im = inputs['image'].cuda()
im_pose = inputs['pose_image'].cuda()
im_h = inputs['head'].cuda()
shape = inputs['shape'].cuda()
agnostic = inputs['agnostic'].cuda()
c = inputs['cloth'].cuda()
cm = inputs['cloth_mask'].cuda()
im_c = inputs['parse_cloth'].cuda()
im_g = inputs['grid_image'].cuda()
gridtps, thetatps, Ipersp, gridaffine, thetaaffine = model(agnostic, c)
warped_cloth = F.grid_sample(Ipersp, gridtps, padding_mode='border') #
warped_maskaffine = F.grid_sample(cm, gridaffine, padding_mode='zeros')
warped_mask = F.grid_sample(warped_maskaffine,
gridtps,
padding_mode='zeros')
warped_gridaffine = F.grid_sample(im_g,
gridaffine,
padding_mode='zeros')
warped_grid = F.grid_sample(warped_gridaffine,
gridtps,
padding_mode='zeros')
visuals = [[im_h, shape, im_pose], [c, warped_cloth, im_c],
[warped_grid, (warped_cloth * 0.7 + im * 0.3), im],
[Ipersp, (Ipersp * 0.7 + im * 0.3), warped_gridaffine]]
Lwarp = criterionL1(warped_cloth, im_c)
Lpersp = criterionL1(Ipersp, im_c)
LgicTPS = gicloss(gridtps)
LgicTPS = LgicTPS / (gridtps.shape[0] * gridtps.shape[1] *
gridtps.shape[2])
# shape of grid: N, Hout, Wout,2: N x 5 x 5 x 2
# grid shape: N x 5 x 5 x 2
loss = 0.5 * Lpersp + 0.5 * Lwarp # + 40 * LgicTPS
optimizer.zero_grad()
loss.backward()
optimizer.step()
if (step + 1) % opt.display_count == 0:
board_add_images(board, 'combine', visuals, step + 1)
board.add_scalar('metric', loss.item(), step + 1)
board.add_scalar('0.5*Lpersp', (0.5 * Lpersp).item(), step + 1)
#board.add_scalar('40*LgicTPS', (40*LgicTPS).item(), step+1)
board.add_scalar('0.5*Lwarp', (0.5 * Lwarp).item(), step + 1)
t = time.time() - iter_start_time
# print('step: %8d, time: %.3f, loss: %4f, (40*LgicTPS): %.8f, 0.7*Lpersp: %.6f, 0.3*Lwarp: %.6f' % (step+1, t, loss.item(), (40*LgicTPS).item(), (0.7*Lpersp).item(), (0.3*Lwarp).item()), flush=True)
print(
'step: %8d, time: %.3f, loss: %4f, 0.5*Lpersp: %.6f, 0.5*Lwarp: %.6f'
% (step + 1, t, loss.item(), (0.5 * Lpersp).item(),
(0.5 * Lwarp).item()),
flush=True)
if (step + 1) % opt.save_count == 0:
save_checkpoint(
model,
os.path.join(opt.checkpoint_dir, opt.name,
'step_%06d.pth' % (step + 1)))
def main():
opt = get_opt()
print(opt)
print("Start to train stage: %s, named: %s!" % (opt.stage, opt.name))
# create dataset
train_dataset = CPDataset(opt)
# create dataloader
train_loader = CPDataLoader(opt, train_dataset)
# visualization
if not os.path.exists(opt.tensorboard_dir):
os.makedirs(opt.tensorboard_dir)
board = SummaryWriter(log_dir=os.path.join(opt.tensorboard_dir, opt.name))
# create model & train & save the final checkpoint
if opt.stage == 'GMM':
if opt.model == 'RefinedGMM':
model = RefinedGMM(opt)
elif opt.model == 'OneRefinedGMM':
model = OneRefinedGMM(opt)
else:
raise TypeError()
if not opt.checkpoint == '' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
train_refined_gmm(opt, train_loader, model, board)
save_checkpoint(
model,
os.path.join(opt.checkpoint_dir, opt.name,
'refined_gmm_final.pth'))
elif opt.stage == 'VariGMM':
model = VariGMM(opt)
if not opt.checkpoint == '' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
train_refined_gmm(opt, train_loader, model, board)
save_checkpoint(
model,
os.path.join(opt.checkpoint_dir, opt.name,
'refined_gmm_final.pth'))
elif opt.stage == 'semanticGMM':
model = RefinedGMM(opt)
if not opt.checkpoint == '' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
train_semantic_parsing_refined_gmm(opt, train_loader, model, board)
save_checkpoint(
model,
os.path.join(opt.checkpoint_dir, opt.name,
'refined_gmm_final.pth'))
elif opt.stage == 'no_background_GMM':
model = RefinedGMM(opt)
if not opt.checkpoint == '' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
train_no_background_refined_gmm(opt, train_loader, model, board)
save_checkpoint(
model,
os.path.join(opt.checkpoint_dir, opt.name,
'no_background_refined_gmm_final.pth'))
elif opt.stage == 'TOM':
model = UnetGenerator(25, 4, 6, ngf=64, norm_layer=nn.InstanceNorm2d)
if not opt.checkpoint == '' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
train_tom(opt, train_loader, model, board)
save_checkpoint(
model, os.path.join(opt.checkpoint_dir, opt.name, 'tom_final.pth'))
elif opt.stage == 'DeepTom':
norm_layer = 'instance'
use_dropout = True
with_tanh = False
model = Define_G(25,
4,
64,
'treeresnet',
'instance',
True,
'normal',
0.02,
opt.gpu_ids,
with_tanh=False)
train_deep_tom(opt, train_loader, model, board)
save_checkpoint(
model, os.path.join(opt.checkpoint_dir, opt.name, 'tom_final.pth'))
else:
raise NotImplementedError('Model [%s] is not implemented' % opt.stage)
print('Finished training %s, nameed: %s!' % (opt.stage, opt.name))
def train_refined_gmm(opt, train_loader, model, board):
model.cuda()
model.train()
loss_weight = opt.loss_weight
# if loss_weight > 0.01:
# print("Error")
# assert False
# criterion
warped_criterionL1 = nn.L1Loss()
result_criterionL1 = nn.L1Loss()
point_criterionL1 = nn.L1Loss()
criterionMask = nn.L1Loss()
criterionVGG = VGGLoss()
criterionGram = GramLoss()
rendered_criterionL1 = nn.L1Loss()
center_mask_critetionL1 = nn.L1Loss()
warped_mask_criterionL1 = nn.L1Loss()
# optimizer
optimizer = torch.optim.Adam(model.parameters(),
lr=opt.lr,
betas=(0.5, 0.999))
scheduler = torch.optim.lr_scheduler.LambdaLR(
optimizer,
lr_lambda=lambda step: 1.0 - max(0, step - opt.keep_step) / float(
opt.decay_step + 1))
for step in range(opt.keep_step + opt.decay_step):
iter_start_time = time.time()
inputs = train_loader.next_batch()
im = inputs['image'].cuda()
im_pose = inputs['pose_image'].cuda()
im_h = inputs['head'].cuda()
shape = inputs['shape'].cuda()
densepose_shape = inputs['densepose_shape'].cuda()
agnostic = inputs['agnostic'].cuda()
c = inputs['cloth'].cuda()
cm = inputs['cloth_mask'].cuda()
im_c = inputs['parse_cloth'].cuda()
im_g = inputs['grid_image'].cuda()
parse_cloth_mask = inputs['parse_cloth_mask'].cuda()
target_shape = inputs['target_shape']
c_point_plane = inputs['cloth_points'].cuda()
p_point_plane = inputs['person_points'].cuda()
grid, theta, warped_cloth, outputs = model(agnostic, c)
#warped_cloth = F.grid_sample(c, grid, padding_mode='border')
warped_mask = F.grid_sample(cm, grid, padding_mode='zeros')
warped_grid = F.grid_sample(im_g, grid, padding_mode='zeros')
compute_c_point_plane = compute_grid_point(p_point_plane, grid)
warped_mask_loss = 0
if opt.add_warped_mask_loss:
warped_mask_loss += warped_mask_criterionL1(
warped_mask, target_shape)
c_rendered, m_composite = torch.split(outputs, 3, 1)
c_rendered = F.tanh(c_rendered)
m_composite = F.sigmoid(m_composite)
c_result = warped_cloth * m_composite + c_rendered * (1 - m_composite)
visuals = [[im_h, shape, im_pose], [c, warped_cloth, im_c],
[warped_grid, (warped_cloth + im) * 0.5, im],
[m_composite, (c_result + im) * 0.5, c_result]]
loss_warped_cloth = warped_criterionL1(warped_cloth, im_c)
loss_point = 0
if opt.add_point_loss:
loss_point = point_criterionL1(compute_c_point_plane,
c_point_plane)
loss_c_result = result_criterionL1(c_result, im_c)
loss_mask = criterionMask(m_composite, warped_mask)
loss_vgg = 0
if opt.add_vgg_loss:
loss_vgg = criterionVGG(c_result, im_c)
loss_gram = 0
if opt.add_gram_loss:
loss_gram += criterionGram(c_result, im_c)
loss_render = 0
if opt.add_render_loss:
loss_render += rendered_criterionL1(c_rendered, im_c)
loss_mask_constrain = 0
if opt.add_mask_constrain:
center_mask = m_composite * parse_cloth_mask
ground_mask = torch.ones_like(parse_cloth_mask, dtype=torch.float)
ground_mask = ground_mask * warped_mask * parse_cloth_mask
loss_mask_constrain = center_mask_critetionL1(
center_mask, ground_mask)
#print("long_mask_constrain = ", loss_mask_constrain)
loss_mask_constrain = loss_mask_constrain * opt.mask_constrain_weight
#print("long_mask_constrain = ", loss_mask_constrain)
# print("loss cloth = ", loss_warped_cloth)
# print("loss point = ", loss_point)
# print("loss render = ", loss_render)
# print("loss_c_result = ", loss_c_result)
loss = loss_warped_cloth + loss_weight * loss_point + loss_c_result + loss_mask + loss_vgg + loss_render + loss_mask_constrain + warped_mask_loss + loss_gram
optimizer.zero_grad()
loss.backward()
optimizer.step()
if (step + 1) % opt.display_count == 0:
board_add_images(board, 'combine', visuals, step + 1)
board.add_scalar('metric', loss.item(), step + 1)
t = time.time() - iter_start_time
print('step: %8d, time: %.3f, loss: %4f' %
(step + 1, t, loss.item()),
flush=True)
if (step + 1) % opt.save_count == 0:
save_checkpoint(
model,
os.path.join(opt.checkpoint_dir, opt.name,
'step_%06d.pth' % (step + 1)))
def train_gmm(opt, train_loader, model, board):
model.cuda()
model.train()
# criterion
criterionL1 = nn.L1Loss()
gicloss = GicLoss(opt)
# optimizer
optimizer = torch.optim.Adam(model.parameters(),
lr=opt.lr,
betas=(0.5, 0.999))
scheduler = torch.optim.lr_scheduler.LambdaLR(
optimizer,
lr_lambda=lambda step: 1.0 - max(0, step - opt.keep_step) / float(
opt.decay_step + 1))
for step in range(opt.keep_step + opt.decay_step):
iter_start_time = time.time()
inputs = train_loader.next_batch()
im = inputs['image'].cuda()
im_pose = inputs['pose_image'].cuda()
im_h = inputs['head'].cuda()
shape = inputs['shape'].cuda()
agnostic = inputs['agnostic'].cuda()
c = inputs['cloth'].cuda()
cm = inputs['cloth_mask'].cuda()
im_c = inputs['parse_cloth'].cuda()
im_g = inputs['grid_image'].cuda()
grid, theta = model(agnostic,
cm) # can be added c too for new training
warped_cloth = F.grid_sample(c, grid, padding_mode='border')
warped_mask = F.grid_sample(cm, grid, padding_mode='zeros')
warped_grid = F.grid_sample(im_g, grid, padding_mode='zeros')
visuals = [[im_h, shape, im_pose], [c, warped_cloth, im_c],
[warped_grid, (warped_cloth + im) * 0.5, im]]
Lwarp = criterionL1(warped_cloth, im_c) # loss for warped cloth
# grid regularization loss
Lgic = gicloss(grid)
# 200x200 = 40.000 * 0.001
Lgic = Lgic / (grid.shape[0] * grid.shape[1] * grid.shape[2])
loss = Lwarp + 40 * Lgic # total GMM loss
optimizer.zero_grad()
loss.backward()
optimizer.step()
if (step + 1) % opt.display_count == 0:
board_add_images(board, 'combine', visuals, step + 1)
board.add_scalar('loss', loss.item(), step + 1)
board.add_scalar('40*Lgic', (40 * Lgic).item(), step + 1)
board.add_scalar('Lwarp', Lwarp.item(), step + 1)
t = time.time() - iter_start_time
print(
'step: %8d, time: %.3f, loss: %4f, (40*Lgic): %.8f, Lwarp: %.6f'
% (step + 1, t, loss.item(), (40 * Lgic).item(), Lwarp.item()),
flush=True)
if (step + 1) % opt.save_count == 0:
save_checkpoint(
model,
os.path.join(opt.checkpoint_dir, opt.name,
'step_%06d.pth' % (step + 1)))
def train_gmm(opt, train_loader, model, board):
model.cuda()
model.train()
# criterion
criterionL1 = nn.L1Loss()
gicloss = GicLoss(opt)
# optimizer
optimizer = torch.optim.Adam(model.parameters(),
lr=opt.lr,
betas=(0.5, 0.999))
scheduler = torch.optim.lr_scheduler.LambdaLR(
optimizer,
lr_lambda=lambda step: 1.0 - max(0, step - opt.keep_step) / float(
opt.decay_step + 1))
for step in range(opt.keep_step + opt.decay_step):
iter_start_time = time.time()
inputs = train_loader.next_batch()
cloth_mask = inputs['cloth_mask'].cuda()
cloth_mask_gt = inputs['cloth_mask_gt'].cuda()
shape = inputs['shape'].cuda()
c_name = inputs['c_name']
im_names = inputs['im_names']
im_namet = inputs['im_namet']
im_g = inputs['grid_image'].cuda()
parse_im = inputs['image'].cuda()
cloth = inputs['c'].cuda()
grid, theta = model(cloth_mask, shape)
warped_mask = F.grid_sample(cloth_mask, grid, padding_mode='zeros')
warped_grid = F.grid_sample(im_g, grid, padding_mode='zeros')
warped_cloth = F.grid_sample(cloth, grid, padding_mode='zeros')
visuals = [[im_g, cloth_mask * 2 - 1, shape],
[cloth, warped_cloth, parse_im],
[warped_grid, warped_mask * 2 - 1, cloth_mask_gt * 2 - 1]]
Lwarp = criterionL1(warped_mask, cloth_mask_gt)
Lgic = gicloss(grid)
Lgic = Lgic / (grid.shape[0] * grid.shape[1] * grid.shape[2]
) #200x200 = 40.000 * 0.001
loss = 0.0244 * Lwarp + 0.9756 * Lgic
optimizer.zero_grad()
loss.backward()
optimizer.step()
if (step + 1) % opt.display_count == 0:
board_add_images(board, 'combine', visuals, step + 1)
board.add_scalar('metric', loss.item(), step + 1)
board.add_scalar('0.9756*Lgic', (0.9756 * Lgic).item(), step + 1)
board.add_scalar('0.0244*Lwarp', (0.0244 * Lwarp).item(), step + 1)
t = time.time() - iter_start_time
print(
'step: %8d, time: %.3f, loss: %4f, (0.9756*Lgic): %.6f, 0.0244*Lwarp: %.4f'
% (step + 1, t, loss.item(), (0.9756 * Lgic).item(),
(0.0244 * Lwarp).item()),
flush=True)
if (step + 1) % opt.save_count == 0:
save_checkpoint(
model,
os.path.join(opt.checkpoint_dir, opt.name,
'step_%06d.pth' % (step + 1)))
def train_residual_old(opt,
train_loader,
model,
model_module,
gmm_model,
generator,
image_embedder,
board,
discriminator=None,
discriminator_module=None):
lambdas_vis_reg = {'l1': 1.0, 'prc': 0.05, 'style': 100.0}
lambdas = {
'adv': 0.1,
'identity': 1000,
'match_gt': 50,
'vis_reg': .1,
'consist': 50
}
model.train()
gmm_model.eval()
image_embedder.eval()
generator.eval()
# criterion
l1_criterion = nn.L1Loss()
mse_criterion = nn.MSELoss()
vgg_extractor = VGGExtractor().cuda().eval()
adv_criterion = utils.AdversarialLoss('lsgan').cuda()
# optimizer
optimizer = torch.optim.Adam(model.parameters(),
lr=opt.lr,
betas=(0.5, 0.999),
weight_decay=1e-4)
if opt.use_gan:
D_optim = torch.optim.Adam(discriminator.parameters(),
lr=opt.lr,
betas=(0.5, 0.999),
weight_decay=1e-4)
pbar = range(opt.keep_step + opt.decay_step)
if single_gpu_flag(opt):
pbar = tqdm(pbar)
for step in pbar:
iter_start_time = time.time()
inputs, inputs_2 = train_loader.next_batch()
im = inputs['image'].cuda()
im_pose = inputs['pose_image']
im_h = inputs['head']
shape = inputs['shape']
agnostic = inputs['agnostic'].cuda()
c = inputs['cloth'].cuda()
cm = inputs['cloth_mask'].cuda()
c_2 = inputs_2['cloth'].cuda()
cm_2 = inputs_2['cloth_mask'].cuda()
with torch.no_grad():
grid, theta = gmm_model(agnostic, c)
c = F.grid_sample(c, grid, padding_mode='border')
cm = F.grid_sample(cm, grid, padding_mode='zeros')
grid_2, theta_2 = gmm_model(agnostic, c_2)
c_2 = F.grid_sample(c_2, grid_2, padding_mode='border')
cm_2 = F.grid_sample(cm_2, grid_2, padding_mode='zeros')
outputs = generator(torch.cat([agnostic, c], 1))
p_rendered, m_composite = torch.split(outputs, 3, 1)
p_rendered = F.tanh(p_rendered)
m_composite = F.sigmoid(m_composite)
transfer_1 = c * m_composite + p_rendered * (1 - m_composite)
outputs_2 = generator(torch.cat([agnostic, c_2], 1))
p_rendered_2, m_composite_2 = torch.split(outputs_2, 3, 1)
p_rendered_2 = F.tanh(p_rendered_2)
m_composite_2 = F.sigmoid(m_composite_2)
transfer_2 = c_2 * m_composite_2 + p_rendered_2 * (1 -
m_composite_2)
gt_residual = (torch.mean(im, dim=1) -
torch.mean(transfer_1, dim=1)).unsqueeze(1)
output_1 = model(transfer_1.detach(), gt_residual.detach())
output_2 = model(transfer_2.detach(), gt_residual.detach())
embedding_1 = image_embedder(output_1)
embedding_2 = image_embedder(output_2)
embedding_1_t = image_embedder(transfer_1)
embedding_2_t = image_embedder(transfer_2)
if opt.use_gan:
# train discriminator
real_L_logit, real_L_cam_logit, real_G_logit, real_G_cam_logit = discriminator(
im)
fake_L_logit_1, fake_L_cam_logit_1, fake_G_logit_1, fake_G_cam_logit_1 = discriminator(
output_1.detach())
fake_L_logit_2, fake_L_cam_logit_2, fake_G_logit_2, fake_G_cam_logit_2 = discriminator(
output_2.detach())
D_true_loss = adv_criterion(real_L_logit, True) + \
adv_criterion(real_G_logit, True) + \
adv_criterion(real_L_cam_logit, True) + \
adv_criterion(real_G_cam_logit, True)
D_fake_loss = adv_criterion(torch.cat([fake_L_cam_logit_1, fake_L_cam_logit_2], dim=0), False) + \
adv_criterion(torch.cat([fake_G_cam_logit_1, fake_G_cam_logit_2], dim=0), False) + \
adv_criterion(torch.cat([fake_L_logit_1, fake_L_logit_2], dim=0), False) + \
adv_criterion(torch.cat([fake_G_logit_1, fake_G_logit_2], dim=0), False)
D_loss = D_true_loss + D_fake_loss
D_optim.zero_grad()
D_loss.backward()
D_optim.step()
# train generator
fake_L_logit_1, fake_L_cam_logit_1, fake_G_logit_1, fake_G_cam_logit_1 = discriminator(
output_1)
fake_L_logit_2, fake_L_cam_logit_2, fake_G_logit_2, fake_G_cam_logit_2 = discriminator(
output_2)
G_adv_loss = adv_criterion(torch.cat([fake_L_logit_1, fake_L_logit_2], dim=0), True) + \
adv_criterion(torch.cat([fake_G_logit_1, fake_G_logit_2], dim=0), True) + \
adv_criterion(torch.cat([fake_L_cam_logit_1, fake_L_cam_logit_2], dim=0), True) + \
adv_criterion(torch.cat([fake_G_cam_logit_1, fake_G_cam_logit_2], dim=0), True)
# identity loss
identity_loss = mse_criterion(embedding_1,
embedding_1_t) + mse_criterion(
embedding_2, embedding_2_t)
# vis reg loss
output_1_feats = vgg_extractor(output_1)
transfer_1_feats = vgg_extractor(transfer_1)
output_2_feats = vgg_extractor(output_2)
transfer_2_feats = vgg_extractor(transfer_2)
# gt_feats = vgg_extractor(data['image'].cuda())
style_reg = utils.compute_style_loss(
output_1_feats,
transfer_1_feats, l1_criterion) + utils.compute_style_loss(
output_2_feats, transfer_2_feats, l1_criterion)
perceptual_reg = utils.compute_perceptual_loss(
output_1_feats, transfer_1_feats,
l1_criterion) + utils.compute_perceptual_loss(
output_2_feats, transfer_2_feats, l1_criterion)
l1_reg = l1_criterion(output_1, transfer_1) + l1_criterion(
output_2, transfer_2)
vis_reg_loss = l1_reg * lambdas_vis_reg[
"l1"] + style_reg * lambdas_vis_reg[
"style"] + perceptual_reg * lambdas_vis_reg["prc"]
# match gt loss
match_gt_loss = l1_criterion(
output_1, im
) #* lambdas_vis_reg["l1"] + utils.compute_style_loss(output_1_feats, gt_feats, l1_criterion) * lambdas_vis_reg["style"] + utils.compute_perceptual_loss(output_1_feats, gt_feats, l1_criterion) * lambdas_vis_reg["prc"]
# consistency loss
consistency_loss = l1_criterion(transfer_1 - output_1,
transfer_2 - output_2)
visuals = [[im_h, shape, im],
[
c, c_2,
torch.cat([gt_residual, gt_residual, gt_residual],
dim=1)
], [transfer_1, output_1, (output_1 - transfer_1) / 2],
[transfer_2, output_2, (output_2 - transfer_2) / 2]]
total_loss = lambdas['identity'] * identity_loss + \
lambdas['match_gt'] * match_gt_loss + \
lambdas['vis_reg'] * vis_reg_loss + \
lambdas['consist'] * consistency_loss
if opt.use_gan:
total_loss += lambdas['adv'] * G_adv_loss
optimizer.zero_grad()
total_loss.backward()
optimizer.step()
if single_gpu_flag(opt):
if (step + 1) % opt.display_count == 0:
board_add_images(board, str(step + 1), visuals, step + 1)
board.add_scalar('loss/total', total_loss.item(), step + 1)
board.add_scalar('loss/identity', identity_loss.item(), step + 1)
board.add_scalar('loss/vis_reg', vis_reg_loss.item(), step + 1)
board.add_scalar('loss/match_gt', match_gt_loss.item(), step + 1)
board.add_scalar('loss/consist', consistency_loss.item(), step + 1)
if opt.use_gan:
board.add_scalar('loss/Dadv', D_loss.item(), step + 1)
board.add_scalar('loss/Gadv', G_adv_loss.item(), step + 1)
pbar.set_description(
'step: %8d, loss: %.4f, identity: %.4f, vis_reg: %.4f, match_gt: %.4f, consist: %.4f'
% (step + 1, total_loss.item(), identity_loss.item(),
vis_reg_loss.item(), match_gt_loss.item(),
consistency_loss.item()))
if (step + 1) % opt.save_count == 0 and single_gpu_flag(opt):
save_checkpoint(
model_module,
os.path.join(opt.checkpoint_dir, opt.name,
'step_%06d.pth' % (step + 1)))
if opt.use_gan:
save_checkpoint(
discriminator_module,
os.path.join(opt.checkpoint_dir, opt.name,
'step_disc_%06d.pth' % (step + 1)))
def train_hpm(opt, train_loader, model, d_g, board):
model.cuda()
model.train()
d_g.cuda()
d_g.train()
dis_label = Variable(torch.FloatTensor(opt.batch_size)).cuda()
# criterion
criterionMCE = nn.CrossEntropyLoss()#nn.BCEWithLogitsLoss()
#criterionMSE = nn.MSELoss()
criterionGAN = nn.BCELoss()
# optimizer
optimizerG = torch.optim.Adam(model.parameters(), lr=opt.lr, betas=(0.5, 0.999))
schedulerG = torch.optim.lr_scheduler.LambdaLR(optimizerG, lr_lambda = lambda step: 1.0 -
max(0, step - opt.keep_step) / float(opt.decay_step + 1))
optimizerD = torch.optim.Adam(d_g.parameters(), lr=opt.lr, betas=(0.5, 0.999))
schedulerD = torch.optim.lr_scheduler.LambdaLR(optimizerD, lr_lambda = lambda step: 1.0 -
max(0, step - opt.keep_step) / float(opt.decay_step + 1))
for step in range(opt.keep_step + opt.decay_step):
iter_start_time = time.time()
#prep
inputs = train_loader.next_batch()
im = inputs['image'].cuda()#sz=b*3*256*192
seg_gt = inputs['seg'].cuda()
seg_enc = inputs['seg_enc'].cuda()
agnostic = inputs['agnostic'].cuda()
c = inputs['cloth'].cuda()
batch_size = im.size(0)
optimizerD.zero_grad()
#D_real
dis_label.data.fill_(1.0-random.random()*0.1)
dis_g_output = d_g(seg_enc)
errDg_real = criterionGAN(dis_g_output, dis_label)
errDg_real.backward()
#generate image
segmentation = model(torch.cat([agnostic, c],1))
#D_fake
dis_label.data.fill_(0.0+random.random()*0.1)
dis_g_output = d_g(segmentation.detach())
errDg_fake = criterionGAN(dis_g_output, dis_label)
errDg_fake.backward()
optimizerD.step()
#model_train
optimizerG.zero_grad()
dis_label.data.fill_(1.0-random.random()*0.1)
dis_g_output = d_g(segmentation)
errG_fake = criterionGAN(dis_g_output, dis_label)
loss_mce = criterionMCE(segmentation.view(batch_size, 16, -1), seg_gt.view(batch_size, -1))
loss = errG_fake + loss_mce * opt.alpha
loss.backward()
optimizerG.step()
if (step+1) % opt.display_count == 0:
t = time.time() - iter_start_time
loss_dict = {"GAN":errG_fake.item(), "TOT":loss.item(), "MCE":loss_mce.item(),
"DG":((errDg_fake+errDg_real)/2).item()}
print('step: %d|time: %.3f'%(step+1, t), end="")
for k, v in loss_dict.items():
print('|%s: %.3f'%(k, v), end="")
board.add_scalar(k, v, step+1)
print()
if (step+1) % opt.save_count == 0:
"""
save_checkpoint(model, os.path.join(opt.checkpoint_dir, opt.stage +'_'+ opt.name, 'step_%06d.pth' % (step+1)))
save_checkpoint(d_g, os.path.join(opt.checkpoint_dir, opt.stage +'_'+ opt.name, 'step_%06d_dg.pth' % (step+1)))
"""
save_checkpoint(model, os.path.join(opt.checkpoint_dir, opt.stage +'_'+ opt.name, 'model.pth'))
save_checkpoint(d_g, os.path.join(opt.checkpoint_dir, opt.stage +'_'+ opt.name, 'dg.pth'))
if opt.shuffle :
train_sampler = None
else:
train_sampler = sampler.RandomSampler(train_dataset)
train_loader = DataLoader(
train_dataset, batch_size=opt.batch_size, shuffle=opt.shuffle,
num_workers=opt.workers, pin_memory=True, sampler=train_sampler)
# visualization
if not os.path.exists(opt.tensorboard_dir):
os.makedirs(opt.tensorboard_dir)
board = SummaryWriter(log_dir = os.path.join(opt.tensorboard_dir, opt.name))
# create model & train & save the final checkpoint
if opt.stage == 'GMM':
model = GMM(opt)
if not opt.checkpoint =='' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
train_gmm(opt, train_loader, model, board)
save_checkpoint(model, os.path.join(opt.checkpoint_dir, opt.name, 'gmm_final.pth'))
elif opt.stage == 'TOM':
model = UnetGenerator(25, 4, 6, ngf=64, norm_layer=nn.InstanceNorm2d)
if not opt.checkpoint =='' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
train_tom(opt, train_loader, model, board)
save_checkpoint(model, os.path.join(opt.checkpoint_dir, opt.name, 'tom_final.pth'))
else:
raise NotImplementedError('Model [%s] is not implemented' % opt.stage)
print('Finished training %s, nameed: %s!' % (opt.stage, opt.name))
def train_gmm(opt, train_loader, model, board):
model.cuda()
discriminator = Discriminator()
discriminator.cuda()
model.train()
discriminator.train()
# criterion
criterion = nn.BCELoss()
criterionL1 = nn.L1Loss()
criterionPSC = PSCLoss()
# optimizer
optimizer = torch.optim.Adam(model.parameters(),
lr=opt.lr,
betas=(0.5, 0.999))
optimizer_d = torch.optim.Adam(discriminator.parameters(),
lr=opt.lr,
betas=(0.5, 0.999))
scheduler = torch.optim.lr_scheduler.LambdaLR(
optimizer,
lr_lambda=lambda step: 1.0 - max(0, step - opt.keep_step) / float(
opt.decay_step + 1))
#count = 0
#base_name = os.path.basename(opt.checkpoint)
# save_dir = os.path.join(opt.result_dir, opt.datamode)
# if not os.path.exists(save_dir):
# os.makedirs(save_dir)
# warp_cloth_dir = os.path.join(save_dir, 'warp-cloth')
# if not os.path.exists(warp_cloth_dir):
# os.makedirs(warp_cloth_dir)
for step in range(opt.keep_step + opt.decay_step):
iter_start_time = time.time()
inputs = train_loader.next_batch()
c_names = inputs['c_name']
im = inputs['image'].cuda()
im_pose = inputs['pose_image'].cuda()
im_h = inputs['head'].cuda()
shape = inputs['shape'].cuda()
agnostic = inputs['agnostic'].cuda()
c = inputs['cloth'].cuda()
cm = inputs['cloth_mask'].cuda()
im_c = inputs['parse_cloth'].cuda()
im_g = inputs['grid_image'].cuda()
blank = inputs['blank'].cuda()
grid, theta = model(agnostic, c)
warped_cloth = F.grid_sample(c, grid, padding_mode='border')
warped_mask = F.grid_sample(cm, grid, padding_mode='zeros')
warped_grid = F.grid_sample(im_g, grid, padding_mode='zeros')
#if (count < 14222):
# save_images(warped_cloth, c_names, warp_cloth_dir)
# print(warped_cloth.size()[0])
# count+=warped_cloth.size()[0]
visuals = [[im_h, shape, im_pose], [c, warped_cloth, im_c],
[warped_grid, (warped_cloth + im) * 0.5, im]]
discriminator_train_step(opt.batch_size, discriminator, model,
optimizer_d, criterion, im_c, agnostic, c)
res, loss, lossL1, lossPSC, lossGAN = generator_train_step(
opt.batch_size, discriminator, model, optimizer, criterion,
criterionL1, criterionPSC, blank, im_c, agnostic, c)
if (step + 1) % opt.display_count == 0:
board_add_images(board, 'combine', visuals, step + 1)
board.add_scalar('lossL1', lossL1.item(), step + 1)
board.add_scalar('lossPSC', lossPSC.item(), step + 1)
board.add_scalar('lossGAN', lossGAN.item(), step + 1)
board.add_scalar('loss', loss.item(), step + 1)
t = time.time() - iter_start_time
print('step: %8d, time: %.3f, lossL1: %4f' %
(step + 1, t, lossL1.item()),
flush=True)
print('step: %8d, time: %.3f, lossPSC: %4f' %
(step + 1, t, lossPSC.item()),
flush=True)
print('step: %8d, time: %.3f, lossGAN: %4f' %
(step + 1, t, lossGAN.item()),
flush=True)
print('step: %8d, time: %.3f, loss: %4f' %
(step + 1, t, loss.item()),
flush=True)
if (step + 1) % opt.save_count == 0:
save_checkpoint(
model,
os.path.join(opt.checkpoint_dir, opt.name,
'step_%06d.pth' % (step + 1)))
def main():
opt = get_opt()
print(opt)
print("Start to train stage: %s, named: %s!" % (opt.stage, opt.name))
n_gpu = int(os.environ['WORLD_SIZE']) if 'WORLD_SIZE' in os.environ else 1
opt.distributed = n_gpu > 1
local_rank = opt.local_rank
if opt.distributed:
torch.cuda.set_device(opt.local_rank)
torch.distributed.init_process_group(backend='nccl',
init_method='env://')
synchronize()
# create dataset
train_dataset = CPDataset(opt)
# create dataloader
train_loader = CPDataLoader(opt, train_dataset)
# visualization
if not os.path.exists(opt.tensorboard_dir):
os.makedirs(opt.tensorboard_dir)
board = None
if single_gpu_flag(opt):
board = SummaryWriter(
log_dir=os.path.join(opt.tensorboard_dir, opt.name))
# create model & train & save the final checkpoint
if opt.stage == 'GMM':
model = GMM(opt)
if not opt.checkpoint == '' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
train_gmm(opt, train_loader, model, board)
save_checkpoint(
model, os.path.join(opt.checkpoint_dir, opt.name, 'gmm_final.pth'))
elif opt.stage == 'TOM':
gmm_model = GMM(opt)
load_checkpoint(gmm_model, "checkpoints/gmm_train_new/step_020000.pth")
gmm_model.cuda()
model = UnetGenerator(25, 4, 6, ngf=64, norm_layer=nn.InstanceNorm2d)
model.cuda()
# if opt.distributed:
# model = torch.nn.SyncBatchNorm.convert_sync_batchnorm(model)
if not opt.checkpoint == '' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
model_module = model
if opt.distributed:
model = torch.nn.parallel.DistributedDataParallel(
model,
device_ids=[local_rank],
output_device=local_rank,
find_unused_parameters=True)
model_module = model.module
train_tom(opt, train_loader, model, model_module, gmm_model, board)
if single_gpu_flag(opt):
save_checkpoint(
model_module,
os.path.join(opt.checkpoint_dir, opt.name, 'tom_final.pth'))
elif opt.stage == 'TOM+WARP':
gmm_model = GMM(opt)
gmm_model.cuda()
model = UnetGenerator(25, 4, 6, ngf=64, norm_layer=nn.InstanceNorm2d)
model.cuda()
# if opt.distributed:
# model = torch.nn.SyncBatchNorm.convert_sync_batchnorm(model)
if not opt.checkpoint == '' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
model_module = model
gmm_model_module = gmm_model
if opt.distributed:
model = torch.nn.parallel.DistributedDataParallel(
model,
device_ids=[local_rank],
output_device=local_rank,
find_unused_parameters=True)
model_module = model.module
gmm_model = torch.nn.parallel.DistributedDataParallel(
gmm_model,
device_ids=[local_rank],
output_device=local_rank,
find_unused_parameters=True)
gmm_model_module = gmm_model.module
train_tom_gmm(opt, train_loader, model, model_module, gmm_model,
gmm_model_module, board)
if single_gpu_flag(opt):
save_checkpoint(
model_module,
os.path.join(opt.checkpoint_dir, opt.name, 'tom_final.pth'))
elif opt.stage == "identity":
model = Embedder()
if not opt.checkpoint == '' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
train_identity_embedding(opt, train_loader, model, board)
save_checkpoint(
model, os.path.join(opt.checkpoint_dir, opt.name, 'gmm_final.pth'))
elif opt.stage == 'residual':
gmm_model = GMM(opt)
load_checkpoint(gmm_model, "checkpoints/gmm_train_new/step_020000.pth")
gmm_model.cuda()
generator_model = UnetGenerator(25,
4,
6,
ngf=64,
norm_layer=nn.InstanceNorm2d)
load_checkpoint(generator_model,
"checkpoints/tom_train_new/step_038000.pth")
generator_model.cuda()
embedder_model = Embedder()
load_checkpoint(embedder_model,
"checkpoints/identity_train_64_dim/step_020000.pth")
embedder_model = embedder_model.embedder_b.cuda()
model = UNet(n_channels=4, n_classes=3)
if opt.distributed:
model = torch.nn.SyncBatchNorm.convert_sync_batchnorm(model)
model.apply(utils.weights_init('kaiming'))
model.cuda()
if opt.use_gan:
discriminator = Discriminator()
discriminator.apply(utils.weights_init('gaussian'))
discriminator.cuda()
acc_discriminator = AccDiscriminator()
acc_discriminator.apply(utils.weights_init('gaussian'))
acc_discriminator.cuda()
if not opt.checkpoint == '' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
if opt.use_gan:
load_checkpoint(discriminator,
opt.checkpoint.replace("step_", "step_disc_"))
model_module = model
if opt.use_gan:
discriminator_module = discriminator
acc_discriminator_module = acc_discriminator
if opt.distributed:
model = torch.nn.parallel.DistributedDataParallel(
model,
device_ids=[local_rank],
output_device=local_rank,
find_unused_parameters=True)
model_module = model.module
if opt.use_gan:
discriminator = torch.nn.parallel.DistributedDataParallel(
discriminator,
device_ids=[local_rank],
output_device=local_rank,
find_unused_parameters=True)
discriminator_module = discriminator.module
acc_discriminator = torch.nn.parallel.DistributedDataParallel(
acc_discriminator,
device_ids=[local_rank],
output_device=local_rank,
find_unused_parameters=True)
acc_discriminator_module = acc_discriminator.module
if opt.use_gan:
train_residual(opt,
train_loader,
model,
model_module,
gmm_model,
generator_model,
embedder_model,
board,
discriminator=discriminator,
discriminator_module=discriminator_module,
acc_discriminator=acc_discriminator,
acc_discriminator_module=acc_discriminator_module)
if single_gpu_flag(opt):
save_checkpoint(
{
"generator": model_module,
"discriminator": discriminator_module
},
os.path.join(opt.checkpoint_dir, opt.name,
'tom_final.pth'))
else:
train_residual(opt, train_loader, model, model_module, gmm_model,
generator_model, embedder_model, board)
if single_gpu_flag(opt):
save_checkpoint(
model_module,
os.path.join(opt.checkpoint_dir, opt.name,
'tom_final.pth'))
elif opt.stage == "residual_old":
gmm_model = GMM(opt)
load_checkpoint(gmm_model, "checkpoints/gmm_train_new/step_020000.pth")
gmm_model.cuda()
generator_model = UnetGenerator(25,
4,
6,
ngf=64,
norm_layer=nn.InstanceNorm2d)
load_checkpoint(generator_model,
"checkpoints/tom_train_new_2/step_070000.pth")
generator_model.cuda()
embedder_model = Embedder()
load_checkpoint(embedder_model,
"checkpoints/identity_train_64_dim/step_020000.pth")
embedder_model = embedder_model.embedder_b.cuda()
model = UNet(n_channels=4, n_classes=3)
if opt.distributed:
model = torch.nn.SyncBatchNorm.convert_sync_batchnorm(model)
model.apply(utils.weights_init('kaiming'))
model.cuda()
if opt.use_gan:
discriminator = Discriminator()
discriminator.apply(utils.weights_init('gaussian'))
discriminator.cuda()
if not opt.checkpoint == '' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
model_module = model
if opt.use_gan:
discriminator_module = discriminator
if opt.distributed:
model = torch.nn.parallel.DistributedDataParallel(
model,
device_ids=[local_rank],
output_device=local_rank,
find_unused_parameters=True)
model_module = model.module
if opt.use_gan:
discriminator = torch.nn.parallel.DistributedDataParallel(
discriminator,
device_ids=[local_rank],
output_device=local_rank,
find_unused_parameters=True)
discriminator_module = discriminator.module
if opt.use_gan:
train_residual_old(opt,
train_loader,
model,
model_module,
gmm_model,
generator_model,
embedder_model,
board,
discriminator=discriminator,
discriminator_module=discriminator_module)
if single_gpu_flag(opt):
save_checkpoint(
{
"generator": model_module,
"discriminator": discriminator_module
},
os.path.join(opt.checkpoint_dir, opt.name,
'tom_final.pth'))
else:
train_residual_old(opt, train_loader, model, model_module,
gmm_model, generator_model, embedder_model,
board)
if single_gpu_flag(opt):
save_checkpoint(
model_module,
os.path.join(opt.checkpoint_dir, opt.name,
'tom_final.pth'))
else:
raise NotImplementedError('Model [%s] is not implemented' % opt.stage)
print('Finished training %s, nameed: %s!' % (opt.stage, opt.name))
def main():
opt = get_opt()
if opt.mode == 'test':
opt.datamode = "test"
opt.data_list = "test_pairs.txt"
opt.shuffle = False
elif opt.mode == 'val':
opt.shuffle = False
elif opt.mode != 'train':
print(opt.mode)
print(opt)
if opt.mode != 'train':
opt.batch_size = 1
if opt.mode != 'train' and not opt.checkpoint:
print("You need to have a checkpoint for: "+opt.mode)
return None
print("Start to train stage: %s, named: %s!" % (opt.stage, opt.name))
# create dataset
train_dataset = CPDataset(opt)
# create dataloader
train_loader = CPDataLoader(opt, train_dataset)
# visualization
if not os.path.exists(opt.tensorboard_dir):
os.makedirs(opt.tensorboard_dir)
board = SummaryWriter(log_dir = os.path.join(opt.tensorboard_dir, opt.name))
# create model & train & save the final checkpoint
if opt.stage == 'HPM':
model = UnetGenerator(25, 16, 6, ngf=64, norm_layer=nn.InstanceNorm2d, clsf=True)
d_g= Discriminator_G(opt, 16)
if not opt.checkpoint =='' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
load_checkpoint(d_g, opt.checkpoint[:-9] + "dg.pth")
if opt.mode == "train":
train_hpm(opt, train_loader, model, d_g, board)
else:
test_hpm(opt, train_loader, model)
save_checkpoints(model, d_g, os.path.join(opt.checkpoint_dir, opt.stage +'_'+ opt.name+"_final", '%s.pth'))
elif opt.stage == 'GMM':
#seg_unet = UnetGenerator(25, 16, 6, ngf=64, norm_layer=nn.InstanceNorm2d, clsf=True)
model = GMM(opt, 1)
if not opt.checkpoint =='' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
if opt.mode == "train":
train_gmm(opt, train_loader, model, board)
else:
test_gmm(opt, train_loader, model)
save_checkpoint(model, os.path.join(opt.checkpoint_dir, opt.name, 'gmm_final.pth'))
elif opt.stage == 'TOM':
model = UnetGenerator(31, 4, 6, ngf=64, norm_layer=nn.InstanceNorm2d)
if not opt.checkpoint =='' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
if opt.mode == "train":
train_tom(opt, train_loader, model, board)
else:
test_tom(opt, train_loader, model)
save_checkpoint(model, os.path.join(opt.checkpoint_dir, opt.name, 'tom_final.pth'))
else:
raise NotImplementedError('Model [%s] is not implemented' % opt.stage)
print('Finished training %s, nameed: %s!' % (opt.stage, opt.name))
def train_gmm(opt, train_loader, model, board):
# load model
model.cuda()
model.train()
# criterion
criterionL1 = nn.L1Loss()
# optimizer
optimizer = torch.optim.Adam(model.parameters(), lr=opt.lr, betas=(0.5, 0.999))
scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer, lr_lambda = lambda step: 1.0 -
max(0, step - opt.keep_step) / float(opt.decay_step + 1))
# train log
if not opt.checkpoint == '':
train_log = open(os.path.join(opt.checkpoint_dir, opt.name, 'train_log.txt'), 'a')
else:
os.makedirs(os.path.join(opt.checkpoint_dir, opt.name), exist_ok=True)
train_log = open(os.path.join(opt.checkpoint_dir, opt.name, 'train_log.txt'), 'w')
train_log.write('='*30 + ' Training Option ' + '='*30 + '\n')
train_log.write(str(opt) + '\n\n')
train_log.write('='*30 + ' Network Architecture ' + '='*30 + '\n')
print(str(model) + '\n', file=train_log)
train_log.write('='*30 + ' Training Log ' + '='*30 + '\n')
# train loop
checkpoint_step = 0
if not opt.checkpoint == '':
checkpoint_step += int(opt.checkpoint.split('/')[-1][5:11])
for step in range(checkpoint_step, opt.keep_step + opt.decay_step):
iter_start_time = time.time()
dl_iter = iter(train_loader)
inputs = dl_iter.next()
im = inputs['image'].cuda()
im_pose = inputs['pose_image'].cuda()
im_h = inputs['head'].cuda()
shape = inputs['shape'].cuda()
agnostic = inputs['agnostic'].cuda()
c = inputs['cloth'].cuda()
cm = inputs['cloth_mask'].cuda()
im_c = inputs['parse_cloth'].cuda()
im_g = inputs['grid_image'].cuda()
grid, theta = model(agnostic, c)
warped_cloth = F.grid_sample(c, grid, padding_mode='border', align_corners=True)
warped_mask = F.grid_sample(cm, grid, padding_mode='zeros', align_corners=True)
warped_grid = F.grid_sample(im_g, grid, padding_mode='zeros', align_corners=True)
visuals = [ [im_h, shape, im_pose],
[c, warped_cloth, im_c],
[warped_grid, (warped_cloth+im)*0.5, im]]
loss = criterionL1(warped_cloth, im_c)
optimizer.zero_grad()
loss.backward()
optimizer.step()
scheduler.step()
if (step+1) % opt.display_count == 0:
board_add_images(board, 'combine', visuals, step+1)
board.add_scalar('metric', loss.item(), step+1)
t = time.time() - iter_start_time
print('step: %8d, time: %.3f, loss: %4f' % (step+1, t, loss.item()), flush=True)
train_log.write('step: %8d, time: %.3f, loss: %.4f'
% (step+1, t, loss.item()) + '\n')
if (step+1) % opt.save_count == 0:
save_checkpoint(model, os.path.join(opt.checkpoint_dir, opt.name, 'step_%06d.pth' % (step+1)))
def train(opt, train_loader, G, D, board):
human_parser = HumanParser(opt)
human_parser.eval()
G.train()
D.train()
# palette = get_palette()
# Criterion
criterionWarp = nn.L1Loss()
criterionPerceptual = VGGLoss()
criterionL1 = nn.L1Loss()
BCE_stable = nn.BCEWithLogitsLoss()
criterionCloth = nn.L1Loss()
# Variables
ya = torch.FloatTensor(opt.batch_size)
yb = torch.FloatTensor(opt.batch_size)
u = torch.FloatTensor((opt.batch_size, 1, 1, 1))
grad_outputs = torch.ones(opt.batch_size)
# Everything cuda
if opt.cuda:
G.cuda()
D.cuda()
human_parser.cuda()
criterionWarp = criterionWarp.cuda()
criterionPerceptual = criterionPerceptual.cuda()
criterionL1 = criterionL1.cuda()
BCE_stable.cuda()
criterionCloth = criterionCloth.cuda()
ya = ya.cuda()
yb = yb.cuda()
u = u.cuda()
grad_outputs = grad_outputs.cuda()
# DataParallel
G = nn.DataParallel(G)
D = nn.DataParallel(D)
human_parser = nn.DataParallel(human_parser)
# Optimizers
optimizerD = torch.optim.Adam(D.parameters(),
lr=opt.lr,
betas=(0.5, 0.999))
optimizerG = torch.optim.Adam(G.parameters(),
lr=opt.lr,
betas=(0.5, 0.999))
# Fitting model
step_start_time = time.time()
for step in range(opt.n_iter):
########################
# (1) Update D network #
########################
for p in D.parameters():
p.requires_grad = True
for t in range(opt.Diters):
D.zero_grad()
inputs = train_loader.next_batch()
pa = inputs['image'].cuda()
ap = inputs['agnostic'].cuda()
cb = inputs['another_cloth'].cuda()
del inputs
current_batch_size = pa.size(0)
ya_pred = D(pa)
_, pb_fake = G(cb, ap)
# Detach y_pred_fake from the neural network G and put it inside D
yb_pred_fake = D(pb_fake.detach())
ya.data.resize_(current_batch_size).fill_(1)
yb.data.resize_(current_batch_size).fill_(0)
errD = (BCE_stable(ya_pred - torch.mean(yb_pred_fake), ya) +
BCE_stable(yb_pred_fake - torch.mean(ya_pred), yb)) / 2.0
errD.backward()
# Gradient penalty
with torch.no_grad():
u.resize_(current_batch_size, 1, 1, 1).uniform_(0, 1)
grad_outputs.data.resize_(current_batch_size)
x_both = pa * u + pb_fake * (1. - u)
# We only want the gradients with respect to x_both
x_both = Variable(x_both, requires_grad=True)
grad = torch.autograd.grad(outputs=D(x_both),
inputs=x_both,
grad_outputs=grad_outputs,
retain_graph=True,
create_graph=True,
only_inputs=True)[0]
# We need to norm 3 times (over n_colors x image_size x image_size) to get only a vector of size
# "batch_size"
grad_penalty = opt.penalty * (
(grad.norm(2, 1).norm(2, 1).norm(2, 1) - 1)**2).mean()
grad_penalty.backward()
optimizerD.step()
########################
# (2) Update G network #
########################
for p in D.parameters():
p.requires_grad = False
for t in range(opt.Giters):
inputs = train_loader.next_batch()
pa = inputs['image'].cuda()
ap = inputs['agnostic'].cuda()
ca = inputs['cloth'].cuda()
cb = inputs['another_cloth'].cuda()
parse_cloth = inputs['parse_cloth'].cuda()
del inputs
current_batch_size = pa.size(0)
# paired data
G.zero_grad()
warped_cloth_a, pa_fake = G(ca, ap)
if step >= opt.human_parser_step: # 生成的图片较真实后再添加human parser
parse_pa_fake = human_parser(pa_fake) # (N,H,W)
parse_ca_fake = (parse_pa_fake == 5) + \
(parse_pa_fake == 6) + \
(parse_pa_fake == 7) # [0,1] (N,H,W)
parse_ca_fake = parse_ca_fake.unsqueeze(1).type_as(
pa_fake) # (N,1,H,W)
ca_fake = pa_fake * parse_ca_fake + (1 - parse_ca_fake
) # [-1,1]
with torch.no_grad():
parse_pa_fake_vis = visualize_seg(parse_pa_fake)
l_cloth_p = criterionCloth(ca_fake, warped_cloth_a)
else:
with torch.no_grad():
ca_fake = torch.zeros_like(pa_fake)
parse_pa_fake_vis = torch.zeros_like(pa_fake)
l_cloth_p = torch.zeros(1).cuda()
l_warp = 20 * criterionWarp(warped_cloth_a, parse_cloth)
l_perceptual = criterionPerceptual(pa_fake, pa)
l_L1 = criterionL1(pa_fake, pa)
loss_p = l_warp + l_perceptual + l_L1 + l_cloth_p
loss_p.backward()
optimizerG.step()
# unpaired data
G.zero_grad()
warped_cloth_b, pb_fake = G(cb, ap)
if step >= opt.human_parser_step: # 生成的图片较真实后再添加human parser
parse_pb_fake = human_parser(pb_fake)
parse_cb_fake = (parse_pb_fake == 5) + \
(parse_pb_fake == 6) + \
(parse_pb_fake == 7) # [0,1] (N,H,W)
parse_cb_fake = parse_cb_fake.unsqueeze(1).type_as(
pb_fake) # (N,1,H,W)
cb_fake = pb_fake * parse_cb_fake + (1 - parse_cb_fake
) # [-1,1]
with torch.no_grad():
parse_pb_fake_vis = visualize_seg(parse_pb_fake)
l_cloth_up = criterionCloth(cb_fake, warped_cloth_b)
else:
with torch.no_grad():
cb_fake = torch.zeros_like(pb_fake)
parse_pb_fake_vis = torch.zeros_like(pb_fake)
l_cloth_up = torch.zeros(1).cuda()
with torch.no_grad():
ya.data.resize_(current_batch_size).fill_(1)
yb.data.resize_(current_batch_size).fill_(0)
ya_pred = D(pa)
yb_pred_fake = D(pb_fake)
# Non-saturating
l_adv = 0.1 * (
BCE_stable(ya_pred - torch.mean(yb_pred_fake), yb) +
BCE_stable(yb_pred_fake - torch.mean(ya_pred), ya)) / 2
loss_up = l_adv + l_cloth_up
loss_up.backward()
optimizerG.step()
# visuals = [
# [cb, warped_cloth_b, pb_fake],
# [ca, warped_cloth_a, pa_fake],
# [ap, parse_cloth, pa]
# ]
visuals = [[
cb, warped_cloth_b, pb_fake, cb_fake, parse_pb_fake_vis
], [ca, warped_cloth_a, pa_fake, ca_fake, parse_pa_fake_vis],
[ap, parse_cloth, pa]]
if (step + 1) % opt.display_count == 0:
board_add_images(board, 'combine', visuals, step + 1)
board.add_scalar('loss_p', loss_p.item(), step + 1)
board.add_scalar('l_warp', l_warp.item(), step + 1)
board.add_scalar('l_perceptual', l_perceptual.item(), step + 1)
board.add_scalar('l_L1', l_L1.item(), step + 1)
board.add_scalar('l_cloth_p', l_cloth_p.item(), step + 1)
board.add_scalar('loss_up', loss_up.item(), step + 1)
board.add_scalar('l_adv', l_adv.item(), step + 1)
board.add_scalar('l_cloth_up', l_cloth_up.item(), step + 1)
board.add_scalar('errD', errD.item(), step + 1)
t = time.time() - step_start_time
print(
'step: %8d, time: %.3f, loss_p: %4f, loss_up: %.4f, l_adv: %.4f, errD: %.4f'
% (step + 1, t, loss_p.item(), loss_up.item(),
l_adv.item(), errD.item()),
flush=True)
step_start_time = time.time()
if (step + 1) % opt.save_count == 0:
save_checkpoint(
G,
os.path.join(opt.checkpoint_dir, opt.name,
'step_%06d.pth' % (step + 1)))
def main():
opt = get_opt()
print(opt)
print("Start to train stage: %s, named: %s!" % (opt.stage, opt.name))
n_gpu = int(os.environ['WORLD_SIZE']) if 'WORLD_SIZE' in os.environ else 1
opt.distributed = n_gpu > 1
local_rank = opt.local_rank
if opt.distributed:
torch.cuda.set_device(opt.local_rank)
torch.distributed.init_process_group(backend='nccl',
init_method='env://')
synchronize()
# create dataset
train_dataset = CPDataset(opt)
# create dataloader
train_loader = CPDataLoader(opt, train_dataset)
# visualization
if not os.path.exists(opt.tensorboard_dir):
os.makedirs(opt.tensorboard_dir)
board = None
if single_gpu_flag(opt):
board = SummaryWriter(
log_dir=os.path.join(opt.tensorboard_dir, opt.name))
gmm_model = GMM(opt)
load_checkpoint(gmm_model, "checkpoints/gmm_train_new/step_020000.pth")
gmm_model.cuda()
generator_model = UnetGenerator(25,
4,
6,
ngf=64,
norm_layer=nn.InstanceNorm2d)
load_checkpoint(generator_model,
"checkpoints/tom_train_new_2/step_040000.pth")
generator_model.cuda()
embedder_model = Embedder()
load_checkpoint(embedder_model,
"checkpoints/identity_train_64_dim/step_020000.pth")
embedder_model = embedder_model.embedder_b.cuda()
model = G()
model.apply(utils.weights_init('kaiming'))
model.cuda()
if opt.use_gan:
discriminator = Discriminator()
discriminator.apply(utils.weights_init('gaussian'))
discriminator.cuda()
if not opt.checkpoint == '' and os.path.exists(opt.checkpoint):
load_checkpoint(model, opt.checkpoint)
model_module = model
if opt.use_gan:
discriminator_module = discriminator
if opt.distributed:
model = torch.nn.parallel.DistributedDataParallel(
model,
device_ids=[local_rank],
output_device=local_rank,
find_unused_parameters=True)
model_module = model.module
if opt.use_gan:
discriminator = torch.nn.parallel.DistributedDataParallel(
discriminator,
device_ids=[local_rank],
output_device=local_rank,
find_unused_parameters=True)
discriminator_module = discriminator.module
if opt.use_gan:
train_residual_old(opt,
train_loader,
model,
model_module,
gmm_model,
generator_model,
embedder_model,
board,
discriminator=discriminator,
discriminator_module=discriminator_module)
if single_gpu_flag(opt):
save_checkpoint(
{
"generator": model_module,
"discriminator": discriminator_module
}, os.path.join(opt.checkpoint_dir, opt.name, 'tom_final.pth'))
else:
train_residual_old(opt, train_loader, model, model_module, gmm_model,
generator_model, embedder_model, board)
if single_gpu_flag(opt):
save_checkpoint(
model_module,
os.path.join(opt.checkpoint_dir, opt.name, 'tom_final.pth'))
print('Finished training %s, nameed: %s!' % (opt.stage, opt.name))