def cal_aflw2000_nme(model, image_folder, isShow=False):
types = ('*.jpg', '*.png')
image_path_list = []
for files in types:
image_path_list.extend(glob.glob(os.path.join(image_folder, files)))
total_num = len(image_path_list)
print(">>>一共有{}张测试图片".format(total_num))
nme_list = []
for i, image_path in tqdm(enumerate(image_path_list)):
image_name = image_path.strip().split('/')[-1]
#对输入的图片进行预处理
src_image, crop_image, kpt, tform = process_image(image_path)
image = test_data_preprocess(crop_image)
image_in = F.normalize(image, FLAGS["normalize_mean"],
FLAGS["normalize_std"], False).unsqueeze_(0)
#预测UV位置映射图
pos = model(image_in).detach().cpu().numpy() * 255
if pos is None:
print("------error------")
continue
#从UV位置映射图中抽取68个人脸关键点
pred_kpt = get_landmarks(pos,
tform) #ndarray:(68,3),,得到的人脸关键点是基于未resize图的。
gt_kpt = kpt.T #(68,3)
#将关键点画出来并保存在图中
if isShow:
outputDir = 'Benchmark/AFLW2000_Result'
plot_image = plot_kpt(src_image, pred_kpt) / 255
save_image_path = os.path.join(outputDir, image_name)
io.imsave(save_image_path, np.squeeze(plot_image)[:, :, ::-1])
#计算 nme
minx, maxx = np.min(gt_kpt[:, 0]), np.max(gt_kpt[:, 0])
miny, maxy = np.min(gt_kpt[:, 1]), np.max(gt_kpt[:, 1])
llength = np.sqrt((maxx - minx) * (maxy - miny))
dis = pred_kpt[:, :2] - gt_kpt[:, :2]
dis = np.sqrt(np.sum(np.power(dis, 2), axis=1))
dis = np.mean(dis)
nme = dis / llength
nme_list.append(nme)
return np.mean(nme_list)
def main(data_dir):
origin_img, uv_map_gt, uv_map_predicted = None, None, None
if not os.path.exists(FLAGS['images']):
os.mkdir(FLAGS['images'])
# 1) Create Dataset of 300_WLP & Dataloader.
wlp300 = PRNetDataset(root_dir=data_dir,
transform=transforms.Compose([
ToTensor(),
ToResize((256, 256)),
ToNormalize(FLAGS["normalize_mean"],
FLAGS["normalize_std"])
]))
wlp300_dataloader = DataLoader(dataset=wlp300,
batch_size=FLAGS['batch_size'],
shuffle=True,
num_workers=1)
# 2) Intermediate Processing.
transform_img = transforms.Compose([
#transforms.ToTensor(),
transforms.Normalize(FLAGS["normalize_mean"], FLAGS["normalize_std"])
])
# 3) Create PRNet model.
start_epoch, target_epoch = FLAGS['start_epoch'], FLAGS['target_epoch']
model = ResFCN256(resolution_input=256,
resolution_output=256,
channel=3,
size=16)
# Load the pre-trained weight
if FLAGS['resume'] != "" and os.path.exists(
os.path.join(FLAGS['pretrained'], FLAGS['resume'])):
state = torch.load(os.path.join(FLAGS['pretrained'], FLAGS['resume']))
model.load_state_dict(state['prnet'])
start_epoch = state['start_epoch']
INFO("Load the pre-trained weight! Start from Epoch", start_epoch)
else:
start_epoch = 0
INFO(
"Pre-trained weight cannot load successfully, train from scratch!")
if torch.cuda.device_count() > 1:
model = torch.nn.DataParallel(model)
model.to(FLAGS["device"])
optimizer = torch.optim.Adam(model.parameters(),
lr=FLAGS["lr"],
betas=(0.5, 0.999))
scheduler = torch.optim.lr_scheduler.ExponentialLR(optimizer, gamma=0.99)
stat_loss = SSIM(mask_path=FLAGS["mask_path"], gauss=FLAGS["gauss_kernel"])
loss = WeightMaskLoss(mask_path=FLAGS["mask_path"])
bce_loss = torch.nn.BCEWithLogitsLoss()
bce_loss.to(FLAGS["device"])
#Loss function for adversarial
for ep in range(start_epoch, target_epoch):
bar = tqdm(wlp300_dataloader)
loss_list_G, stat_list = [], []
for i, sample in enumerate(bar):
uv_map, origin = sample['uv_map'].to(
FLAGS['device']), sample['origin'].to(FLAGS['device'])
# Inference.
optimizer.zero_grad()
uv_map_result = model(origin)
loss_g = bce_loss(uv_map_result, uv_map)
loss_g.backward()
loss_list_G.append(loss_g.item())
optimizer.step()
if ep % FLAGS["save_interval"] == 0:
with torch.no_grad():
print(" {} [BCE ({})]".format(ep, loss_list_G[-1]))
origin = cv2.imread("./test_data/obama_origin.jpg")
gt_uv_map = np.load("./test_data/test_obama.npy")
origin, gt_uv_map = test_data_preprocess(
origin), test_data_preprocess(gt_uv_map)
origin, gt_uv_map = transform_img(origin), transform_img(
gt_uv_map)
origin_in = origin.unsqueeze_(0).cuda()
pred_uv_map = model(origin_in).detach().cpu()
save_image(
[origin.cpu(),
gt_uv_map.unsqueeze_(0).cpu(), pred_uv_map],
os.path.join(FLAGS['images'],
str(ep) + '.png'),
nrow=1,
normalize=True)
# Save model
print("Save model")
state = {
'prnet': model.state_dict(),
'Loss': loss_list_G,
'start_epoch': ep
}
torch.save(state, os.path.join(FLAGS['images'],
'{}.pth'.format(ep)))
scheduler.step()
def main(data_dir):
# 0) Tensoboard Writer.
writer = SummaryWriter(FLAGS['summary_path'])
origin_img, uv_map_gt, uv_map_predicted = None, None, None
if not os.path.exists(FLAGS['images']):
os.mkdir(FLAGS['images'])
# 1) Create Dataset of 300_WLP & Dataloader.
wlp300 = PRNetDataset(root_dir=data_dir,
transform=transforms.Compose([ToTensor(),
ToNormalize(FLAGS["normalize_mean"], FLAGS["normalize_std"])]))
wlp300_dataloader = DataLoader(dataset=wlp300, batch_size=FLAGS['batch_size'], shuffle=True, num_workers=4)
# 2) Intermediate Processing.
transform_img = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize(FLAGS["normalize_mean"], FLAGS["normalize_std"])
])
# 3) Create PRNet model.
start_epoch, target_epoch = FLAGS['start_epoch'], FLAGS['target_epoch']
model = ResFCN256()
# Load the pre-trained weight
if FLAGS['resume'] and os.path.exists(os.path.join(FLAGS['images'], "latest.pth")):
state = torch.load(os.path.join(FLAGS['images'], "latest.pth"))
model.load_state_dict(state['prnet'])
start_epoch = state['start_epoch']
INFO("Load the pre-trained weight! Start from Epoch", start_epoch)
else:
start_epoch = 0
INFO("Pre-trained weight cannot load successfully, train from scratch!")
if torch.cuda.device_count() > 1:
model = torch.nn.DataParallel(model)
model.to("cuda")
optimizer = torch.optim.Adam(model.parameters(), lr=FLAGS["lr"], betas=(0.5, 0.999))
scheduler = torch.optim.lr_scheduler.ExponentialLR(optimizer, gamma=0.99)
stat_loss = SSIM(mask_path=FLAGS["mask_path"], gauss=FLAGS["gauss_kernel"])
loss = WeightMaskLoss(mask_path=FLAGS["mask_path"])
for ep in range(start_epoch, target_epoch):
bar = tqdm(wlp300_dataloader)
Loss_list, Stat_list = [], []
for i, sample in enumerate(bar):
uv_map, origin = sample['uv_map'].to(FLAGS['device']), sample['origin'].to(FLAGS['device'])
# Inference.
uv_map_result = model(origin)
# Loss & ssim stat.
logit = loss(uv_map_result, uv_map)
stat_logit = stat_loss(uv_map_result, uv_map)
# Record Loss.
Loss_list.append(logit.item())
Stat_list.append(stat_logit.item())
# Update.
optimizer.zero_grad()
logit.backward()
optimizer.step()
bar.set_description(" {} [Loss(Paper)] {} [SSIM({})] {}".format(ep, Loss_list[-1], FLAGS["gauss_kernel"], Stat_list[-1]))
# Record Training information in Tensorboard.
if origin_img is None and uv_map_gt is None:
origin_img, uv_map_gt = origin, uv_map
uv_map_predicted = uv_map_result
writer.add_scalar("Original Loss", Loss_list[-1], FLAGS["summary_step"])
writer.add_scalar("SSIM Loss", Stat_list[-1], FLAGS["summary_step"])
grid_1, grid_2, grid_3 = make_grid(origin_img, normalize=True), make_grid(uv_map_gt), make_grid(uv_map_predicted)
writer.add_image('original', grid_1, FLAGS["summary_step"])
writer.add_image('gt_uv_map', grid_2, FLAGS["summary_step"])
writer.add_image('predicted_uv_map', grid_3, FLAGS["summary_step"])
writer.add_graph(model, uv_map)
if ep % FLAGS["save_interval"] == 0:
with torch.no_grad():
origin = cv2.imread("./test_data/obama_origin.jpg")
gt_uv_map = np.load("./test_data/test_obama.npy")
origin, gt_uv_map = test_data_preprocess(origin), test_data_preprocess(gt_uv_map)
# origin, gt_uv_map = transform_img(origin), transform_img(gt_uv_map)
origin_in = origin.unsqueeze_(0).cuda()
pred_uv_map = model(origin_in).detach().cpu()
save_image([origin.cpu(), gt_uv_map.unsqueeze_(0).cpu(), pred_uv_map],
os.path.join(FLAGS['images'], str(ep) + '.png'), nrow=1, normalize=True)
# Save model
state = {
'prnet': model.state_dict(),
'Loss': Loss_list,
'start_epoch': ep,
}
torch.save(state, os.path.join(FLAGS['images'], 'latest.pth'))
scheduler.step()
writer.close()
def main(data_dir):
# 0) Tensoboard Writer.
writer = SummaryWriter(FLAGS['summary_path'])
origin_img, uv_map_gt, uv_map_predicted = None, None, None
if not os.path.exists(FLAGS['images']):
os.mkdir(FLAGS['images'])
# 1) Create Dataset of 300_WLP & Dataloader.
wlp300 = PRNetDataset(root_dir=data_dir,
transform=transforms.Compose([
ToTensor(),
ToResize((416, 416)),
ToNormalize(FLAGS["normalize_mean"],
FLAGS["normalize_std"])
]))
wlp300_dataloader = DataLoader(dataset=wlp300,
batch_size=FLAGS['batch_size'],
shuffle=True,
num_workers=1)
# 2) Intermediate Processing.
transform_img = transforms.Compose([
transforms.Normalize(FLAGS["normalize_mean"], FLAGS["normalize_std"])
])
# 3) Create PRNet model.
start_epoch, target_epoch = FLAGS['start_epoch'], FLAGS['target_epoch']
model = ResFCN256(resolution_input=416,
resolution_output=416,
channel=3,
size=16)
discriminator = Discriminator()
# Load the pre-trained weight
if FLAGS['resume'] != "" and os.path.exists(
os.path.join(FLAGS['pretrained'], FLAGS['resume'])):
state = torch.load(os.path.join(FLAGS['pretrained'], FLAGS['resume']))
model.load_state_dict(state['prnet'])
start_epoch = state['start_epoch']
INFO("Load the pre-trained weight! Start from Epoch", start_epoch)
else:
start_epoch = 0
INFO(
"Pre-trained weight cannot load successfully, train from scratch!")
if torch.cuda.device_count() > 1:
model = torch.nn.DataParallel(model)
model.to(FLAGS["device"])
discriminator.to(FLAGS["device"])
optimizer = torch.optim.Adam(model.parameters(),
lr=FLAGS["lr"],
betas=(0.5, 0.999))
optimizer_D = torch.optim.Adam(discriminator.parameters(), lr=FLAGS["lr"])
scheduler = torch.optim.lr_scheduler.ExponentialLR(optimizer, gamma=0.99)
stat_loss = SSIM(mask_path=FLAGS["mask_path"], gauss=FLAGS["gauss_kernel"])
loss = WeightMaskLoss(mask_path=FLAGS["mask_path"])
bce_loss = torch.nn.BCEWithLogitsLoss()
bce_loss.to(FLAGS["device"])
#Loss function for adversarial
for ep in range(start_epoch, target_epoch):
bar = tqdm(wlp300_dataloader)
loss_list_G, stat_list = [], []
loss_list_D = []
for i, sample in enumerate(bar):
uv_map, origin = sample['uv_map'].to(
FLAGS['device']), sample['origin'].to(FLAGS['device'])
# Inference.
optimizer.zero_grad()
uv_map_result = model(origin)
# Update D
optimizer_D.zero_grad()
fake_detach = uv_map_result.detach()
d_fake = discriminator(fake_detach)
d_real = discriminator(uv_map)
retain_graph = False
if FLAGS['gan_type'] == 'GAN':
loss_d = bce_loss(d_real, d_fake)
elif FLAGS['gan_type'].find('WGAN') >= 0:
loss_d = (d_fake - d_real).mean()
if FLAGS['gan_type'].find('GP') >= 0:
epsilon = torch.rand(fake_detach.shape[0]).view(
-1, 1, 1, 1)
epsilon = epsilon.to(fake_detach.device)
hat = fake_detach.mul(1 - epsilon) + uv_map.mul(epsilon)
hat.requires_grad = True
d_hat = discriminator(hat)
gradients = torch.autograd.grad(outputs=d_hat.sum(),
inputs=hat,
retain_graph=True,
create_graph=True,
only_inputs=True)[0]
gradients = gradients.view(gradients.size(0), -1)
gradient_norm = gradients.norm(2, dim=1)
gradient_penalty = 10 * gradient_norm.sub(1).pow(2).mean()
loss_d += gradient_penalty
# from ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks
elif FLAGS['gan_type'] == 'RGAN':
better_real = d_real - d_fake.mean(dim=0, keepdim=True)
better_fake = d_fake - d_real.mean(dim=0, keepdim=True)
loss_d = bce_loss(better_real, better_fake)
retain_graph = True
if discriminator.training:
loss_list_D.append(loss_d.item())
loss_d.backward(retain_graph=retain_graph)
optimizer_D.step()
if 'WGAN' in FLAGS['gan_type']:
for p in discriminator.parameters():
p.data.clamp_(-1, 1)
# Update G
d_fake_bp = discriminator(
uv_map_result) # for backpropagation, use fake as it is
if FLAGS['gan_type'] == 'GAN':
label_real = torch.ones_like(d_fake_bp)
loss_g = bce_loss(d_fake_bp, label_real)
elif FLAGS['gan_type'].find('WGAN') >= 0:
loss_g = -d_fake_bp.mean()
elif FLAGS['gan_type'] == 'RGAN':
better_real = d_real - d_fake_bp.mean(dim=0, keepdim=True)
better_fake = d_fake_bp - d_real.mean(dim=0, keepdim=True)
loss_g = bce_loss(better_fake, better_real)
loss_g.backward()
loss_list_G.append(loss_g.item())
optimizer.step()
stat_logit = stat_loss(uv_map_result, uv_map)
stat_list.append(stat_logit.item())
#bar.set_description(" {} [Loss(Paper)] {} [Loss(D)] {} [SSIM({})] {}".format(ep, loss_list_G[-1], loss_list_D[-1],FLAGS["gauss_kernel"], stat_list[-1]))
# Record Training information in Tensorboard.
"""
if origin_img is None and uv_map_gt is None:
origin_img, uv_map_gt = origin, uv_map
uv_map_predicted = uv_map_result
writer.add_scalar("Original Loss", loss_list_G[-1], FLAGS["summary_step"])
writer.add_scalar("D Loss", loss_list_D[-1], FLAGS["summary_step"])
writer.add_scalar("SSIM Loss", stat_list[-1], FLAGS["summary_step"])
grid_1, grid_2, grid_3 = make_grid(origin_img, normalize=True), make_grid(uv_map_gt), make_grid(uv_map_predicted)
writer.add_image('original', grid_1, FLAGS["summary_step"])
writer.add_image('gt_uv_map', grid_2, FLAGS["summary_step"])
writer.add_image('predicted_uv_map', grid_3, FLAGS["summary_step"])
writer.add_graph(model, uv_map)
"""
if ep % FLAGS["save_interval"] == 0:
with torch.no_grad():
print(" {} [Loss(Paper)] {} [Loss(D)] {} [SSIM({})] {}".format(
ep, loss_list_G[-1], loss_list_D[-1],
FLAGS["gauss_kernel"], stat_list[-1]))
origin = cv2.imread("./test_data/obama_origin.jpg")
gt_uv_map = np.load("./test_data/test_obama.npy")
origin, gt_uv_map = test_data_preprocess(
origin), test_data_preprocess(gt_uv_map)
origin, gt_uv_map = transform_img(origin), transform_img(
gt_uv_map)
origin_in = origin.unsqueeze_(0).cuda()
pred_uv_map = model(origin_in).detach().cpu()
save_image(
[origin.cpu(),
gt_uv_map.unsqueeze_(0).cpu(), pred_uv_map],
os.path.join(FLAGS['images'],
str(ep) + '.png'),
nrow=1,
normalize=True)
# Save model
print("Save model")
state = {
'prnet': model.state_dict(),
'Loss': loss_list_G,
'start_epoch': ep,
'Loss_D': loss_list_D,
}
torch.save(state, os.path.join(FLAGS['images'],
'{}.pth'.format(ep)))
scheduler.step()
writer.close()
def main(data_dir):
# 0) Tensoboard Writer.
writer = SummaryWriter(FLAGS['summary_path'])
origin_img, uv_map_gt, uv_map_predicted = None, None, None
if not os.path.exists(FLAGS['images']):
os.mkdir(FLAGS['images'])
# 1) Create Dataset of 300_WLP & Dataloader.
wlp300 = PRNetDataset(root_dir=data_dir,
transform=transforms.Compose([
ToTensor(),
ToNormalize(FLAGS["normalize_mean"],
FLAGS["normalize_std"])
]))
wlp300_dataloader = DataLoader(dataset=wlp300,
batch_size=FLAGS['batch_size'],
shuffle=True,
num_workers=4)
# 2) Intermediate Processing.
transform_img = transforms.Compose([
#transforms.ToTensor(),
transforms.Normalize(FLAGS["normalize_mean"], FLAGS["normalize_std"])
])
# 3) Create PRNet model.
start_epoch, target_epoch = FLAGS['start_epoch'], FLAGS['target_epoch']
model = ResFCN256()
discriminator = Discriminator1()
# Load the pre-trained weight
if FLAGS['resume'] and os.path.exists(
os.path.join(FLAGS['images'], "latest.pth")):
state = torch.load(os.path.join(FLAGS['images'], "latest.pth"))
model.load_state_dict(state['prnet'])
start_epoch = state['start_epoch']
INFO("Load the pre-trained weight! Start from Epoch", start_epoch)
else:
start_epoch = 0
INFO(
"Pre-trained weight cannot load successfully, train from scratch!")
if torch.cuda.device_count() > 1:
model = torch.nn.DataParallel(model)
model.to("cuda")
discriminator.to("cuda")
optimizer = torch.optim.Adam(model.parameters(),
lr=FLAGS["lr"],
betas=(0.5, 0.999))
optimizer_D = torch.optim.Adam(discriminator.parameters(), lr=FLAGS["lr"])
scheduler = torch.optim.lr_scheduler.ExponentialLR(optimizer, gamma=0.99)
stat_loss = SSIM(mask_path=FLAGS["mask_path"], gauss=FLAGS["gauss_kernel"])
loss = WeightMaskLoss(mask_path=FLAGS["mask_path"])
bce_loss = torch.nn.BCEWithLogitsLoss()
bce_loss.to("cuda")
#Loss function for adversarial
Tensor = torch.cuda.FloatTensor
for ep in range(start_epoch, target_epoch):
bar = tqdm(wlp300_dataloader)
Loss_list, Stat_list = [], []
Loss_list_D = []
for i, sample in enumerate(bar):
#drop the last batch
if i == 111:
break
# the dimension of the uv_map is 16 3 256 256 but the last sample is 10 3 256 256
uv_map, origin = sample['uv_map'].to(
FLAGS['device']), sample['origin'].to(FLAGS['device'])
# generate fake label
fake_label = Variable(torch.zeros([16, 512])).cuda()
# generate real lable
real_label = Variable(torch.ones([16, 512])).cuda()
# Sample noise as generator input
#z = Variable(Tensor(np.random.normal(0, 1,([16, 3, 256, 256]))))
# Inference.
uv_map_result = model(origin)
# Loss & ssim stat.
# Loss measures generator's ability to fool the discriminator
logit = bce_loss(discriminator(uv_map_result), real_label)
stat_logit = stat_loss(uv_map_result, uv_map)
# Measure discriminator's ability to classify real from generated samples
real_loss = bce_loss(discriminator(uv_map), real_label)
fake_loss = bce_loss(
discriminator(uv_map_result).detach(), fake_label)
d_loss = (real_loss + fake_loss) / 2
# Record Loss.
Loss_list.append(logit.item())
Loss_list_D.append(d_loss.item())
Stat_list.append(stat_logit.item())
# Update.
optimizer.zero_grad()
logit.backward()
optimizer.step()
bar.set_description(
" {} [Loss(Paper)] {} [Loss(D)] {} [SSIM({})] {}".format(
ep, Loss_list[-1], Loss_list_D[-1], FLAGS["gauss_kernel"],
Stat_list[-1]))
optimizer_D.zero_grad()
d_loss.backward(retain_graph=True)
optimizer_D.step()
# Record Training information in Tensorboard.
if origin_img is None and uv_map_gt is None:
origin_img, uv_map_gt = origin, uv_map
uv_map_predicted = uv_map_result
writer.add_scalar("Original Loss", Loss_list[-1],
FLAGS["summary_step"])
writer.add_scalar("D Loss", Loss_list_D[-1], FLAGS["summary_step"])
writer.add_scalar("SSIM Loss", Stat_list[-1],
FLAGS["summary_step"])
grid_1, grid_2, grid_3 = make_grid(
origin_img, normalize=True), make_grid(uv_map_gt), make_grid(
uv_map_predicted)
writer.add_image('original', grid_1, FLAGS["summary_step"])
writer.add_image('gt_uv_map', grid_2, FLAGS["summary_step"])
writer.add_image('predicted_uv_map', grid_3, FLAGS["summary_step"])
writer.add_graph(model, uv_map)
if ep % FLAGS["save_interval"] == 0:
with torch.no_grad():
origin = cv2.imread("./test_data/obama_origin.jpg")
gt_uv_map = np.load("./test_data/test_obama.npy")
origin, gt_uv_map = test_data_preprocess(
origin), test_data_preprocess(gt_uv_map)
origin, gt_uv_map = transform_img(origin), transform_img(
gt_uv_map)
origin_in = origin.unsqueeze_(0).cuda()
pred_uv_map = model(origin_in).detach().cpu()
save_image(
[origin.cpu(),
gt_uv_map.unsqueeze_(0).cpu(), pred_uv_map],
os.path.join(FLAGS['images'],
str(ep) + '.png'),
nrow=1,
normalize=True)
# Save model
print("Save model")
state = {
'prnet': model.state_dict(),
'Loss': Loss_list,
'start_epoch': ep,
'Loss_D': Loss_list_D,
}
torch.save(state, os.path.join(FLAGS['images'], 'latest.pth'))
scheduler.step()
writer.close()
def main(data_dir):
origin_img, uv_map_gt, uv_map_predicted = None, None, None
if not os.path.exists(FLAGS['images']):
os.mkdir(FLAGS['images'])
# 1) Create Dataset of 300_WLP & Dataloader.
wlp300 = PRNetDataset(root_dir=data_dir,
transform=transforms.Compose([
ToTensor(),
ToResize((416, 416)),
ToNormalize(FLAGS["normalize_mean"],
FLAGS["normalize_std"])
]))
wlp300_dataloader = DataLoader(dataset=wlp300,
batch_size=FLAGS['batch_size'],
shuffle=True,
num_workers=1)
# 2) Intermediate Processing.
transform_img = transforms.Compose([
#transforms.ToTensor(),
transforms.Normalize(FLAGS["normalize_mean"], FLAGS["normalize_std"])
])
# 3) Create PRNet model.
start_epoch, target_epoch = FLAGS['start_epoch'], FLAGS['target_epoch']
g_x = ResFCN256(resolution_input=416,
resolution_output=416,
channel=3,
size=16)
g_y = ResFCN256(resolution_input=416,
resolution_output=416,
channel=3,
size=16)
d_x = Discriminator()
d_y = Discriminator()
# Load the pre-trained weight
if FLAGS['resume'] != "" and os.path.exists(
os.path.join(FLAGS['pretrained'], FLAGS['resume'])):
state = torch.load(os.path.join(FLAGS['pretrained'], FLAGS['resume']))
try:
g_x.load_state_dict(state['g_x'])
g_y.load_state_dict(state['g_y'])
d_x.load_state_dict(state['d_x'])
d_y.load_state_dict(state['d_y'])
except Exception:
g_x.load_state_dict(state['prnet'])
start_epoch = state['start_epoch']
INFO("Load the pre-trained weight! Start from Epoch", start_epoch)
else:
start_epoch = 0
INFO(
"Pre-trained weight cannot load successfully, train from scratch!")
if torch.cuda.device_count() > 1:
model = torch.nn.DataParallel(model)
g_x.to(FLAGS["device"])
g_y.to(FLAGS["device"])
d_x.to(FLAGS["device"])
d_y.to(FLAGS["device"])
optimizer_g = torch.optim.Adam(itertools.chain(g_x.parameters(),
g_y.parameters()),
lr=FLAGS["lr"],
betas=(0.5, 0.999))
optimizer_d = torch.optim.Adam(itertools.chain(d_x.parameters(),
d_y.parameters()),
lr=FLAGS["lr"])
scheduler = torch.optim.lr_scheduler.ExponentialLR(optimizer_g, gamma=0.99)
stat_loss = SSIM(mask_path=FLAGS["mask_path"], gauss=FLAGS["gauss_kernel"])
loss = WeightMaskLoss(mask_path=FLAGS["mask_path"])
bce_loss = torch.nn.BCEWithLogitsLoss()
bce_loss.to(FLAGS["device"])
l1_loss = nn.L1Loss().to(FLAGS["device"])
lambda_X = 10
lambda_Y = 10
#Loss function for adversarial
for ep in range(start_epoch, target_epoch):
bar = tqdm(wlp300_dataloader)
loss_list_cycle_x = []
loss_list_cycle_y = []
loss_list_d_x = []
loss_list_d_y = []
real_label = torch.ones(FLAGS['batch_size'])
fake_label = torch.zeros(FLAGS['batch_size'])
for i, sample in enumerate(bar):
real_y, real_x = sample['uv_map'].to(
FLAGS['device']), sample['origin'].to(FLAGS['device'])
# x -> y' -> x^
optimizer_g.zero_grad()
fake_y = g_x(real_x)
prediction = d_x(fake_y)
loss_g_x = bce_loss(prediction, real_label)
x_hat = g_y(fake_y)
loss_cycle_x = l1_loss(x_hat, real_x) * lambda_X
loss_x = loss_g_x + loss_cycle_x
loss_x.backward(retain_graph=True)
optimizer_g.step()
loss_list_cycle_x.append(loss_x.item())
# y -> x' -> y^
optimizer_g.zero_grad()
fake_x = g_y(real_y)
prediction = d_y(fake_x)
loss_g_y = bce_loss(prediction, real_label)
y_hat = g_x(fake_x)
loss_cycle_y = l1_loss(y_hat, real_y) * lambda_Y
loss_y = loss_g_y + loss_cycle_y
loss_y.backward(retain_graph=True)
optimizer_g.step()
loss_list_cycle_y.append(loss_y.item())
# d_x
optimizer_d.zero_grad()
pred_real = d_x(real_y)
loss_d_x_real = bce_loss(pred_real, real_label)
pred_fake = d_x(fake_y)
loss_d_x_fake = bce_loss(pred_fake, fake_label)
loss_d_x = (loss_d_x_real + loss_d_x_fake) * 0.5
loss_d_x.backward()
loss_list_d_x.append(loss_d_x.item())
optimizer_d.step()
if 'WGAN' in FLAGS['gan_type']:
for p in d_x.parameters():
p.data.clamp_(-1, 1)
# d_y
optimizer_d.zero_grad()
pred_real = d_y(real_x)
loss_d_y_real = bce_loss(pred_real, real_label)
pred_fake = d_y(fake_x)
loss_d_y_fake = bce_loss(pred_fake, fake_label)
loss_d_y = (loss_d_y_real + loss_d_y_fake) * 0.5
loss_d_y.backward()
loss_list_d_y.append(loss_d_y.item())
optimizer_d.step()
if 'WGAN' in FLAGS['gan_type']:
for p in d_y.parameters():
p.data.clamp_(-1, 1)
if ep % FLAGS["save_interval"] == 0:
with torch.no_grad():
print(
" {} [Loss_G_X] {} [Loss_G_Y] {} [Loss_D_X] {} [Loss_D_Y] {}"
.format(ep, loss_list_g_x[-1], loss_list_g_y[-1],
loss_list_d_x[-1], loss_list_d_y[-1]))
origin = cv2.imread("./test_data/obama_origin.jpg")
gt_uv_map = np.load("./test_data/test_obama.npy")
origin, gt_uv_map = test_data_preprocess(
origin), test_data_preprocess(gt_uv_map)
origin, gt_uv_map = transform_img(origin), transform_img(
gt_uv_map)
origin_in = origin.unsqueeze_(0).cuda()
pred_uv_map = g_x(origin_in).detach().cpu()
save_image(
[origin.cpu(),
gt_uv_map.unsqueeze_(0).cpu(), pred_uv_map],
os.path.join(FLAGS['images'],
str(ep) + '.png'),
nrow=1,
normalize=True)
# Save model
print("Save model")
state = {
'g_x': g_x.state_dict(),
'g_y': g_y.state_dict(),
'd_x': d_x.state_dict(),
'd_y': d_y.state_dict(),
'start_epoch': ep,
}
torch.save(state, os.path.join(FLAGS['images'],
'{}.pth'.format(ep)))
scheduler.step()
def main(data_dir):
# 0) Tensoboard Writer.
writer = SummaryWriter(FLAGS['summary_path'])
origin_img, uv_map_gt, uv_map_predicted = None, None, None
# 1) Create Dataset of 300_WLP.
train_data_dir = [
'/home/beitadoge/Github/PRNet_PyTorch/Data/PRNet_PyTorch_Data/300WLP_AFW_HELEN_LFPW',
'/home/beitadoge/Github/PRNet_PyTorch/Data/PRNet_PyTorch_Data/300WLP_AFW_HELEN_LFPW_Flip',
'/home/beitadoge/Github/PRNet_PyTorch/Data/PRNet_PyTorch_Data/300WLP_IBUG_Src_Flip'
]
wlp300 = PRNetDataset(root_dir=train_data_dir,
transform=transforms.Compose([
ToTensor(),
ToNormalize(FLAGS["normalize_mean"],
FLAGS["normalize_std"])
]))
# 2) Create DataLoader.
wlp300_dataloader = DataLoaderX(dataset=wlp300,
batch_size=FLAGS['batch_size'],
shuffle=True,
num_workers=4)
# 3) Create PRNet model.
start_epoch, target_epoch = FLAGS['start_epoch'], FLAGS['target_epoch']
model = ResFCN256()
#GPU
if torch.cuda.device_count() > 1:
model = torch.nn.DataParallel(model)
model.to("cuda")
#Optimizer
optimizer = torch.optim.Adam(model.parameters(),
lr=FLAGS["lr"],
betas=(0.5, 0.999))
# scheduler_MultiStepLR = torch.optim.lr_scheduler.MultiStepLR(optimizer,[11],gamma=0.5, last_epoch=-1)
scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(optimizer,
mode='min',
factor=0.1,
patience=5,
min_lr=1e-6,
verbose=False)
# scheduler = torch.optim.lr_scheduler.ExponentialLR(optimizer, gamma=0.99)
# scheduler_StepLR = torch.optim.lr_scheduler.StepLR(optimizer,step_size=5,gamma=0.5, last_epoch=-1)
#apex混合精度训练
# from apex import amp
# model , optimizer = amp.initialize(model,optimizer,opt_level="O1",verbosity=0)
#Loss
stat_loss = SSIM(mask_path=FLAGS["mask_path"], gauss=FLAGS["gauss_kernel"])
loss = WeightMaskLoss(mask_path=FLAGS["mask_path"])
# Load the pre-trained weight
if FLAGS['resume'] and os.path.exists(
os.path.join(FLAGS['model_path'], "latest.pth")):
state = torch.load(os.path.join(
FLAGS['model_path'],
"latest.pth")) #这是个字典,keys: ['prnet', 'Loss', 'start_epoch']
model.load_state_dict(state['prnet'])
optimizer.load_state_dict(state['optimizer'])
# amp.load_state_dict(state['amp'])
start_epoch = state['start_epoch']
INFO("Load the pre-trained weight! Start from Epoch", start_epoch)
else:
start_epoch = 0
INFO(
"Pre-trained weight cannot load successfully, train from scratch!")
#Tensorboard
model_input = torch.rand(FLAGS['batch_size'], 3, 256, 256)
writer.add_graph = (model, model_input)
nme_mean = 999
for ep in range(start_epoch, target_epoch):
bar = tqdm(wlp300_dataloader)
Loss_list, Stat_list = deque(maxlen=len(bar)), deque(maxlen=len(bar))
model.train()
for i, sample in enumerate(bar):
uv_map, origin = sample['uv_map'].to(
FLAGS['device']), sample['origin'].to(FLAGS['device'])
# Inference.
uv_map_result = model(origin)
# Loss & ssim stat.
logit_loss = loss(uv_map_result, uv_map)
stat_logit = stat_loss(uv_map_result, uv_map)
# Record Loss.
Loss_list.append(logit_loss.item())
Stat_list.append(stat_logit.item())
# Update.
optimizer.zero_grad()
logit_loss.backward()
# with amp.scale_loss(logit_loss,optimizer) as scaled_loss:
# scaled_loss.backward()
optimizer.step()
lr = optimizer.param_groups[0]['lr']
bar.set_description(
" {} lr {} [Loss(Paper)] {:.5f} [SSIM({})] {:.5f}".format(
ep, lr, np.mean(Loss_list), FLAGS["gauss_kernel"],
np.mean(Stat_list)))
# Record Training information in Tensorboard.
# if origin_img is None and uv_map_gt is None:
# origin_img, uv_map_gt = origin, uv_map
# uv_map_predicted = uv_map_result
#写入Tensorboard
# FLAGS["summary_step"] += 1
# if FLAGS["summary_step"] % 500 ==0:
# writer.add_scalar("Original Loss", Loss_list[-1], FLAGS["summary_step"])
# writer.add_scalar("SSIM Loss", Stat_list[-1], FLAGS["summary_step"])
# grid_1, grid_2, grid_3 = make_grid(origin_img, normalize=True), make_grid(uv_map_gt), make_grid(uv_map_predicted)
# writer.add_image('original', grid_1, FLAGS["summary_step"])
# writer.add_image('gt_uv_map', grid_2, FLAGS["summary_step"])
# writer.add_image('predicted_uv_map', grid_3, FLAGS["summary_step"])
# writer.add_graph(model, uv_map)
#每个epoch过后将Loss写入Tensorboard
loss_mean = np.mean(Loss_list)
writer.add_scalar("Original Loss", loss_mean, ep)
lr = optimizer.param_groups[0]['lr']
writer.add_scalar("lr", lr, ep)
# scheduler_StepLR.step()
scheduler.step(loss_mean)
del Loss_list
del Stat_list
#Test && Cal AFLW2000's NME
model.eval()
if ep % FLAGS["save_interval"] == 0:
with torch.no_grad():
nme_mean = cal_aflw2000_nme(
model, '/home/beitadoge/Data/PRNet_PyTorch_Data/AFLW2000')
print("NME IS {}".format(nme_mean))
writer.add_scalar("Aflw2000_nme", nme_mean, ep)
origin = cv2.imread("./test_data/obama_origin.jpg")
gt_uv_map = cv2.imread("./test_data/obama_uv_posmap.jpg")
origin, gt_uv_map = test_data_preprocess(
origin), test_data_preprocess(gt_uv_map)
origin_in = F.normalize(origin, FLAGS["normalize_mean"],
FLAGS["normalize_std"],
False).unsqueeze_(0)
pred_uv_map = model(origin_in).detach().cpu()
save_image(
[origin.cpu(),
gt_uv_map.unsqueeze_(0).cpu(), pred_uv_map],
os.path.join(FLAGS['model_path'],
str(ep) + '.png'),
nrow=1,
normalize=True)
# # Save model
# state = {
# 'prnet': model.state_dict(),
# 'Loss': Loss_list,
# 'start_epoch': ep,
# }
# torch.save(checkpoint, os.path.join(FLAGS['model_path'], 'epoch{}.pth'.format(ep)))
checkpoint = {
'model': model.state_dict(),
'optimizer': optimizer.state_dict(),
# 'amp': amp.state_dict(),
'start_epoch': ep
}
torch.save(checkpoint,
os.path.join(FLAGS['model_path'], 'lastest.pth'))
# adjust_learning_rate(lr , ep, optimizer)
# scheduler.step(nme_mean)
writer.close()