def __init__(self, args):
self.args = args
self.device = torch.device(args.device)
# image transform
input_transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize([.485, .456, .406], [.229, .224, .225]),
])
# dataset and dataloader
val_dataset = get_segmentation_dataset('eyes',
split='val',
mode='testval',
transform=input_transform)
val_sampler = make_data_sampler(val_dataset, False, args.distributed)
val_batch_sampler = make_batch_data_sampler(val_sampler,
images_per_batch=1)
self.val_loader = data.DataLoader(dataset=val_dataset,
batch_sampler=val_batch_sampler,
num_workers=args.workers,
pin_memory=True)
# create network
self.model = get_segmentation_model(model=args.model,
dataset=args.dataset,
aux=args.aux,
pretrained=True,
pretrained_base=False)
if args.distributed:
self.model = self.model.module
self.model.to(self.device)
self.metric = SegmentationMetric(val_dataset.num_class)
def __init__(self, args):
self.args = args
self.device = torch.device(args.device)
# image transform
input_transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize([.485, .456, .406], [.229, .224, .225]),
])
# dataset and dataloader
val60_data_kwargs = {
'transform': input_transform,
'base_size': args.base_size,
'crop_size': args.crop_size,
're_size': args.re_size,
}
valset = get_segmentation_dataset(args.dataset,
args=args,
split='val',
mode='val_onlyrs',
**val60_data_kwargs)
val_sampler = make_data_sampler(valset, True, args.distributed)
val_batch_sampler = make_batch_data_sampler(val_sampler,
args.batch_size)
self.val60_loader = data.DataLoader(dataset=valset,
batch_sampler=val_batch_sampler,
num_workers=args.workers,
pin_memory=True)
# create network
BatchNorm2d = nn.SyncBatchNorm if args.distributed else nn.BatchNorm2d
self.model = get_segmentation_model(args.model,
dataset=args.dataset,
args=self.args,
norm_layer=BatchNorm2d).to(
self.device)
self.model = load_model(args.resume, self.model)
# evaluation metrics
self.metric_120 = SegmentationMetric(valset.num_class)
self.metric_60 = SegmentationMetric(valset.num_class)
self.best_pred = 0.0
def __init__(self, args):
self.args = args
self.device = torch.device(args.device)
self.nb_classes = 3
valid_img_dir = '/home/wangjialei/teeth_dataset/new_data_20190621/valid_new/images'
valid_mask_dir = '/home/wangjialei/teeth_dataset/new_data_20190621/valid_new/masks'
# valid_transform=transforms.Compose([
# # transforms.ToTensor(),
# # transforms.Normalize([0.517446, 0.360147, 0.310427], [0.061526,0.049087, 0.041330])#R_var is 0.061526, G_var is 0.049087, B_var is 0.041330
# ])
# dataset and dataloader
valid_set = SegmentationData(images_dir=valid_img_dir,
masks_dir=valid_mask_dir,
nb_classes=self.nb_classes,
mode='valid',
transform=None)
valid_sampler = make_data_sampler(valid_set, False, args.distributed)
valid_batch_sampler = make_batch_data_sampler(valid_sampler,
images_per_batch=1)
self.val_loader = data.DataLoader(dataset=valid_set,
batch_sampler=valid_batch_sampler,
num_workers=args.workers,
pin_memory=True)
# create network
self.model = get_segmentation_model(model=args.model,
dataset=args.dataset,
aux=args.aux,
pretrained=True,
pretrained_base=False)
if args.distributed:
self.model = self.model.module
self.model.to(self.device)
self.metric = SegmentationMetric(valid_set.num_class)
def __init__(self, args):
self.args = args
self.device = torch.device(args.device)
# image transform
input_transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize([.485, .456, .406], [.229, .224, .225]),
])
# dataset and dataloader
data_kwargs = {
'transform': input_transform,
'base_size': args.base_size,
'crop_size': args.crop_size
}
trainset = get_segmentation_dataset(args.dataset,
split='train',
mode='train',
**data_kwargs)
args.iters_per_epoch = len(trainset) // (args.num_gpus *
args.batch_size)
args.max_iters = args.epochs * args.iters_per_epoch
train_sampler = make_data_sampler(trainset,
shuffle=True,
distributed=args.distributed)
train_batch_sampler = make_batch_data_sampler(train_sampler,
args.batch_size,
args.max_iters)
self.train_loader = data.DataLoader(dataset=trainset,
batch_sampler=train_batch_sampler,
num_workers=args.workers,
pin_memory=True)
if not args.skip_val:
valset = get_segmentation_dataset(args.dataset,
split='val',
mode='val',
**data_kwargs)
val_sampler = make_data_sampler(valset, False, args.distributed)
val_batch_sampler = make_batch_data_sampler(
val_sampler, args.batch_size)
self.val_loader = data.DataLoader(dataset=valset,
batch_sampler=val_batch_sampler,
num_workers=args.workers,
pin_memory=True)
# create network
BatchNorm2d = nn.SyncBatchNorm if args.distributed else nn.BatchNorm2d
self.model = get_segmentation_model(args.model,
dataset=args.dataset,
aux=args.aux,
norm_layer=BatchNorm2d)
if args.distributed:
self.model = nn.parallel.DistributedDataParallel(
self.model,
device_ids=[args.local_rank],
output_device=args.local_rank)
self.model = self.model.to(args.device)
# resume checkpoint if needed
if args.resume:
if os.path.isfile(args.resume):
name, ext = os.path.splitext(args.resume)
assert ext == '.pkl' or '.pth', 'Sorry only .pth and .pkl files supported.'
print('Resuming training, loading {}...'.format(args.resume))
self.model.load_state_dict(
torch.load(args.resume,
map_location=lambda storage, loc: storage))
# create criterion
if args.ohem:
min_kept = int(args.batch_size // args.num_gpus *
args.crop_size**2 // 16)
self.criterion = MixSoftmaxCrossEntropyOHEMLoss(
args.aux, args.aux_weight, min_kept=min_kept,
ignore_index=-1).to(self.device)
else:
self.criterion = MixSoftmaxCrossEntropyLoss(args.aux,
args.aux_weight,
ignore_index=-1).to(
self.device)
# optimizer
self.optimizer = torch.optim.SGD(self.model.parameters(),
lr=args.lr,
momentum=args.momentum,
weight_decay=args.weight_decay)
# lr scheduling
self.lr_scheduler = WarmupPolyLR(self.optimizer,
max_iters=args.max_iters,
power=0.9,
warmup_factor=args.warmup_factor,
warmup_iters=args.warmup_iters,
warmup_method=args.warmup_method)
# evaluation metrics
self.metric = SegmentationMetric(trainset.num_class)
self.best_pred = 0.0
class Trainer(object):
def __init__(self, args):
self.args = args
self.device = torch.device(args.device)
# image transform
input_transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize([.485, .456, .406], [.229, .224, .225]),
])
# dataset and dataloader
data_kwargs = {
'transform': input_transform,
'base_size': args.base_size,
'crop_size': args.crop_size
}
trainset = get_segmentation_dataset(args.dataset,
split='train',
mode='train',
**data_kwargs)
args.iters_per_epoch = len(trainset) // (args.num_gpus *
args.batch_size)
args.max_iters = args.epochs * args.iters_per_epoch
train_sampler = make_data_sampler(trainset,
shuffle=True,
distributed=args.distributed)
train_batch_sampler = make_batch_data_sampler(train_sampler,
args.batch_size,
args.max_iters)
self.train_loader = data.DataLoader(dataset=trainset,
batch_sampler=train_batch_sampler,
num_workers=args.workers,
pin_memory=True)
if not args.skip_val:
valset = get_segmentation_dataset(args.dataset,
split='val',
mode='val',
**data_kwargs)
val_sampler = make_data_sampler(valset, False, args.distributed)
val_batch_sampler = make_batch_data_sampler(
val_sampler, args.batch_size)
self.val_loader = data.DataLoader(dataset=valset,
batch_sampler=val_batch_sampler,
num_workers=args.workers,
pin_memory=True)
# create network
BatchNorm2d = nn.SyncBatchNorm if args.distributed else nn.BatchNorm2d
self.model = get_segmentation_model(args.model,
dataset=args.dataset,
aux=args.aux,
norm_layer=BatchNorm2d)
if args.distributed:
self.model = nn.parallel.DistributedDataParallel(
self.model,
device_ids=[args.local_rank],
output_device=args.local_rank)
self.model = self.model.to(args.device)
# resume checkpoint if needed
if args.resume:
if os.path.isfile(args.resume):
name, ext = os.path.splitext(args.resume)
assert ext == '.pkl' or '.pth', 'Sorry only .pth and .pkl files supported.'
print('Resuming training, loading {}...'.format(args.resume))
self.model.load_state_dict(
torch.load(args.resume,
map_location=lambda storage, loc: storage))
# create criterion
if args.ohem:
min_kept = int(args.batch_size // args.num_gpus *
args.crop_size**2 // 16)
self.criterion = MixSoftmaxCrossEntropyOHEMLoss(
args.aux, args.aux_weight, min_kept=min_kept,
ignore_index=-1).to(self.device)
else:
self.criterion = MixSoftmaxCrossEntropyLoss(args.aux,
args.aux_weight,
ignore_index=-1).to(
self.device)
# optimizer
self.optimizer = torch.optim.SGD(self.model.parameters(),
lr=args.lr,
momentum=args.momentum,
weight_decay=args.weight_decay)
# lr scheduling
self.lr_scheduler = WarmupPolyLR(self.optimizer,
max_iters=args.max_iters,
power=0.9,
warmup_factor=args.warmup_factor,
warmup_iters=args.warmup_iters,
warmup_method=args.warmup_method)
# evaluation metrics
self.metric = SegmentationMetric(trainset.num_class)
self.best_pred = 0.0
def train(self):
save_to_disk = get_rank() == 0
epochs, max_iters = self.args.epochs, self.args.max_iters
log_per_iters, val_per_iters = self.args.log_iter, self.args.val_epoch * self.args.iters_per_epoch
save_per_iters = self.args.save_epoch * self.args.iters_per_epoch
start_time = time.time()
logger.info(
'Start training, Total Epochs: {:d} = Total Iterations {:d}'.
format(epochs, max_iters))
self.model.train()
for iteration, (images, targets) in enumerate(self.train_loader):
iteration += 1
self.lr_scheduler.step()
images = images.to(self.device)
targets = targets.to(self.device)
outputs = self.model(images)
loss_dict = self.criterion(outputs, targets)
losses = sum(loss for loss in loss_dict.values())
# reduce losses over all GPUs for logging purposes
loss_dict_reduced = reduce_loss_dict(loss_dict)
losses_reduced = sum(loss for loss in loss_dict_reduced.values())
self.optimizer.zero_grad()
losses.backward()
self.optimizer.step()
eta_seconds = ((time.time() - start_time) /
iteration) * (max_iters - iteration)
eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))
if iteration % log_per_iters == 0 and save_to_disk:
logger.info(
"Iters: {:d}/{:d} || Lr: {:.6f} || Loss: {:.4f} || Cost Time: {} || Estimated Time: {}"
.format(
iteration, max_iters,
self.optimizer.param_groups[0]['lr'],
losses_reduced.item(),
str(
datetime.timedelta(seconds=int(time.time() -
start_time))),
eta_string))
if iteration % save_per_iters == 0 and save_to_disk:
save_checkpoint(self.model, self.args, is_best=False)
if not self.args.skip_val and iteration % val_per_iters == 0:
self.validation()
self.model.train()
save_checkpoint(self.model, self.args, is_best=False)
total_training_time = time.time() - start_time
total_training_str = str(
datetime.timedelta(seconds=total_training_time))
logger.info("Total training time: {} ({:.4f}s / it)".format(
total_training_str, total_training_time / max_iters))
def validation(self):
# total_inter, total_union, total_correct, total_label = 0, 0, 0, 0
is_best = False
self.metric.reset()
if self.args.distributed:
model = self.model.module
else:
model = self.model
torch.cuda.empty_cache() # TODO check if it helps
model.eval()
for i, (image, target) in enumerate(self.val_loader):
image = image.to(self.device)
target = target.to(self.device)
with torch.no_grad():
outputs = model(image)
self.metric.update(outputs[0], target)
pixAcc, mIoU = self.metric.get()
logger.info(
"Sample: {:d}, Validation pixAcc: {:.3f}, mIoU: {:.3f}".format(
i + 1, pixAcc, mIoU))
new_pred = (pixAcc + mIoU) / 2
if new_pred > self.best_pred:
is_best = True
self.best_pred = new_pred
save_checkpoint(self.model, self.args, is_best)
synchronize()
def __init__(self, args):
self.args = args
self.device = torch.device(args.device)
# image transform
input_transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize([.485, .456, .406], [.229, .224, .225]),
])
# dataset and dataloader
train_data_kwargs = {
'transform': input_transform,
'base_size': args.base_size,
'crop_size': args.crop_size,
're_size': args.re_size,
}
trainset = get_segmentation_dataset(args.dataset,
args=args,
split='train',
mode='train_onlyrs',
**train_data_kwargs)
args.iters_per_epoch = len(trainset) // (args.num_gpus *
args.batch_size)
args.max_iters = args.epochs * args.iters_per_epoch
train_sampler = make_data_sampler(trainset,
shuffle=True,
distributed=args.distributed)
train_batch_sampler = make_batch_data_sampler(train_sampler,
args.batch_size,
args.max_iters)
self.train_loader = data.DataLoader(dataset=trainset,
batch_sampler=train_batch_sampler,
num_workers=args.workers,
pin_memory=True)
val60_data_kwargs = {
'transform': input_transform,
'base_size': args.base_size,
'crop_size': args.crop_size,
're_size': args.re_size,
}
valset = get_segmentation_dataset(args.dataset,
args=args,
split='val',
mode='val_onlyrs',
**val60_data_kwargs)
val_sampler = make_data_sampler(valset, True, args.distributed)
val_batch_sampler = make_batch_data_sampler(val_sampler,
args.batch_size)
self.val60_loader = data.DataLoader(dataset=valset,
batch_sampler=val_batch_sampler,
num_workers=args.workers,
pin_memory=True)
# create network
BatchNorm2d = nn.SyncBatchNorm if args.distributed else nn.BatchNorm2d
self.model = get_segmentation_model(args.model,
dataset=args.dataset,
args=self.args,
norm_layer=BatchNorm2d).to(
self.device)
self.model = load_modules(args, self.model)
self.model = fix_model(args, self.model)
# optimizer
self.optimizer = torch.optim.SGD(filter(lambda p: p.requires_grad,
self.model.parameters()),
lr=args.lr,
momentum=args.momentum,
weight_decay=args.weight_decay)
# create criterion
if args.ohem:
min_kept = int(args.batch_size // args.num_gpus *
args.crop_size**2 // 16)
self.criterion = MixSoftmaxCrossEntropyOHEMLoss(
args.aux, args.aux_weight, min_kept=min_kept,
ignore_index=-1).to(self.device)
else:
self.criterion = MixSoftmaxCrossEntropyLoss(args.aux,
args.aux_weight,
ignore_index=-1).to(
self.device)
# lr scheduling
self.lr_scheduler = WarmupPolyLR(self.optimizer,
max_iters=args.max_iters,
power=0.9,
warmup_factor=args.warmup_factor,
warmup_iters=args.warmup_iters,
warmup_method=args.warmup_method)
if args.use_DataParallel:
self.model = torch.nn.DataParallel(self.model,
device_ids=range(
torch.cuda.device_count()))
elif args.distributed:
self.model = nn.parallel.DistributedDataParallel(
self.model,
device_ids=[args.local_rank],
output_device=args.local_rank,
find_unused_parameters=True)
# evaluation metrics
self.metric_120 = SegmentationMetric(trainset.num_class)
self.metric_60 = SegmentationMetric(trainset.num_class)
self.best_pred = 0.0
class Trainer(object):
def __init__(self, args):
self.args = args
self.device = torch.device(args.device)
# image transform
input_transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize([.485, .456, .406], [.229, .224, .225]),
])
# dataset and dataloader
train_data_kwargs = {
'transform': input_transform,
'base_size': args.base_size,
'crop_size': args.crop_size,
're_size': args.re_size,
}
trainset = get_segmentation_dataset(args.dataset,
args=args,
split='train',
mode='train_onlyrs',
**train_data_kwargs)
args.iters_per_epoch = len(trainset) // (args.num_gpus *
args.batch_size)
args.max_iters = args.epochs * args.iters_per_epoch
train_sampler = make_data_sampler(trainset,
shuffle=True,
distributed=args.distributed)
train_batch_sampler = make_batch_data_sampler(train_sampler,
args.batch_size,
args.max_iters)
self.train_loader = data.DataLoader(dataset=trainset,
batch_sampler=train_batch_sampler,
num_workers=args.workers,
pin_memory=True)
val60_data_kwargs = {
'transform': input_transform,
'base_size': args.base_size,
'crop_size': args.crop_size,
're_size': args.re_size,
}
valset = get_segmentation_dataset(args.dataset,
args=args,
split='val',
mode='val_onlyrs',
**val60_data_kwargs)
val_sampler = make_data_sampler(valset, True, args.distributed)
val_batch_sampler = make_batch_data_sampler(val_sampler,
args.batch_size)
self.val60_loader = data.DataLoader(dataset=valset,
batch_sampler=val_batch_sampler,
num_workers=args.workers,
pin_memory=True)
# create network
BatchNorm2d = nn.SyncBatchNorm if args.distributed else nn.BatchNorm2d
self.model = get_segmentation_model(args.model,
dataset=args.dataset,
args=self.args,
norm_layer=BatchNorm2d).to(
self.device)
self.model = load_modules(args, self.model)
self.model = fix_model(args, self.model)
# optimizer
self.optimizer = torch.optim.SGD(filter(lambda p: p.requires_grad,
self.model.parameters()),
lr=args.lr,
momentum=args.momentum,
weight_decay=args.weight_decay)
# create criterion
if args.ohem:
min_kept = int(args.batch_size // args.num_gpus *
args.crop_size**2 // 16)
self.criterion = MixSoftmaxCrossEntropyOHEMLoss(
args.aux, args.aux_weight, min_kept=min_kept,
ignore_index=-1).to(self.device)
else:
self.criterion = MixSoftmaxCrossEntropyLoss(args.aux,
args.aux_weight,
ignore_index=-1).to(
self.device)
# lr scheduling
self.lr_scheduler = WarmupPolyLR(self.optimizer,
max_iters=args.max_iters,
power=0.9,
warmup_factor=args.warmup_factor,
warmup_iters=args.warmup_iters,
warmup_method=args.warmup_method)
if args.use_DataParallel:
self.model = torch.nn.DataParallel(self.model,
device_ids=range(
torch.cuda.device_count()))
elif args.distributed:
self.model = nn.parallel.DistributedDataParallel(
self.model,
device_ids=[args.local_rank],
output_device=args.local_rank,
find_unused_parameters=True)
# evaluation metrics
self.metric_120 = SegmentationMetric(trainset.num_class)
self.metric_60 = SegmentationMetric(trainset.num_class)
self.best_pred = 0.0
def train(self, writer):
save_to_disk = get_rank() == 0
epochs, max_iters = self.args.epochs, self.args.max_iters
log_per_iters, val_per_iters = self.args.log_iter, self.args.val_epoch * self.args.iters_per_epoch
save_per_iters = self.args.save_epoch * self.args.iters_per_epoch
start_time = time.time()
logger.info(
'Start training, Total Epochs: {:d} = Total Iterations {:d}'.
format(epochs, max_iters))
self.model.train()
for iteration, (images, targets, _) in enumerate(self.train_loader):
iteration += self.args.start_step
self.lr_scheduler.step()
for index in range(len(images)):
images[index] = images[index].to(self.device)
for index in range(len(targets)):
targets[index] = targets[index].to(self.device)
outputs = self.model(images)
loss_dict = self.criterion(outputs, targets)
losses = sum(loss for loss in loss_dict.values())
# reduce losses over all GPUs for logging purposes
loss_dict_reduced = reduce_loss_dict(loss_dict)
losses_reduced = sum(loss for loss in loss_dict_reduced.values())
self.optimizer.zero_grad()
losses.backward()
self.optimizer.step()
eta_seconds = ((time.time() - start_time) /
iteration) * (max_iters - iteration)
eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))
writer.add_scalar("learning_rate",
self.optimizer.param_groups[0]['lr'], iteration)
writer.add_scalar("Loss/train_loss", losses_reduced.item(),
iteration)
if iteration % log_per_iters == 0 and save_to_disk:
logger.info(
"Iters: {:d}/{:d} || Lr: {:.6f} || Loss: {:.4f} || Estimated Time: {}"
.format(iteration, max_iters,
self.optimizer.param_groups[0]['lr'],
losses_reduced.item(), eta_string))
if iteration % save_per_iters == 0 and save_to_disk:
print('saving......')
save_checkpoint(self.model,
self.args,
iteration=iteration,
is_best=False)
print('save over!')
if (iteration % val_per_iters == 0):
print('evaluating...')
self.validate(iteration, writer)
self.model.train()
print('eval over!')
total_training_time = time.time() - start_time
total_training_str = str(
datetime.timedelta(seconds=total_training_time))
logger.info("Total training time: {} ({:.4f}s / it)".format(
total_training_str, total_training_time / max_iters))
def validate(self, iteration, writer):
# total_inter, total_union, total_correct, total_label = 0, 0, 0, 0
is_best = False
self.metric_120.reset()
self.metric_60.reset()
if self.args.distributed:
model = self.model.module
else:
model = self.model
torch.cuda.empty_cache() # TODO check if it helps
model.eval()
loss = [[], []]
for i, (image, target, _) in enumerate(self.val60_loader):
for index in range(len(image)):
image[index] = image[index].to(self.device)
for index in range(len(target)):
target[index] = target[index].to(self.device)
with torch.no_grad():
outputs = model(image)
self.metric_120.update(outputs[0][0], target[0])
self.metric_60.update(outputs[0][1], target[1])
loss_dict = self.criterion(outputs, target)
loss_dict_120 = loss_dict['loss_120']
loss_dict_60 = loss_dict['loss_60']
loss_dict_reduced_120 = reduce_loss_dict(loss_dict_120)
loss_dict_reduced_60 = reduce_loss_dict(loss_dict_60)
loss[0].append(loss_dict_reduced_120)
loss[1].append(loss_dict_reduced_60)
pixAcc_120, mIoU_120, Iou_120 = self.metric_120.get()
val_loss_120 = sum(loss[0]) / len(loss[0])
val_mIou_120 = mIoU_120
val_mpixAcc_120 = pixAcc_120
logger.info(
"120 Loss: {:.3f}, Validation mpixAcc: {:.3f}, mIoU: {:.3f}".
format(val_loss_120, val_mpixAcc_120, val_mIou_120))
writer.add_scalar("Loss/val120_loss", val_loss_120, iteration)
writer.add_scalar("Result/val120_mIou", val_mIou_120, iteration)
writer.add_scalar("Result/val120_Acc", val_mpixAcc_120, iteration)
for i, j in enumerate(Iou_120):
logger.info("class {:d} : {:.3f}".format(i, j))
writer.add_scalar("Class120/class_{}".format(i), Iou_120[i],
iteration)
pixAcc_60, mIoU_60, Iou_60 = self.metric_60.get()
val_loss_60 = sum(loss[1]) / len(loss[1])
val_mIou_60 = mIoU_60
val_mpixAcc_60 = pixAcc_60
logger.info(
"60 Loss: {:.3f}, Validation mpixAcc: {:.3f}, mIoU: {:.3f}".
format(val_loss_60, val_mpixAcc_60, val_mIou_60))
writer.add_scalar("Loss/val60_loss", val_loss_60, iteration)
writer.add_scalar("Result/val60_mIou", val_mIou_60, iteration)
writer.add_scalar("Result/val60_Acc", val_mpixAcc_60, iteration)
for i, j in enumerate(Iou_60):
logger.info("class {:d} : {:.3f}".format(i, j))
writer.add_scalar("Class60/class_{}".format(i), Iou_60[i],
iteration)
new_pred = (val_mIou_60 + val_mIou_120) / 2.0
if new_pred > self.best_pred:
is_best = True
self.best_pred = new_pred
save_checkpoint(self.model, self.args, iteration, is_best)
synchronize()
def __init__(self, args):
self.args = args
self.device = torch.device(args.device)
self.nb_classes = 3
train_img_dir = '/home/wangjialei/teeth_dataset/new_data_20190621/train_new/images'
train_mask_dir = '/home/wangjialei/teeth_dataset/new_data_20190621/train_new/masks'
# image transform
train_transform=transforms.Compose([
transforms.ToTensor(),
transforms.Normalize([0.519401, 0.359217, 0.310136], [0.061113, 0.048637, 0.041166]),#R_var is 0.061113, G_var is 0.048637, B_var is 0.041166
])
# dataset and dataloader
data_kwargs = {'transform': train_transform, 'base_size': args.base_size, 'crop_size': args.crop_size}
# train_set = get_segmentation_dataset(args.dataset, train_img_dir, train_mask_dir, self.nb_classes, 'train', **data_kwargs)
train_set = SegmentationData(images_dir=train_img_dir, masks_dir=train_mask_dir, nb_classes=self.nb_classes, mode='train', transform=train_transform)
args.iters_per_epoch = len(train_set) // (args.num_gpus * args.batch_size)
args.max_iters = args.epochs * args.iters_per_epoch
train_sampler = make_data_sampler(train_set, shuffle=True, distributed=args.distributed)
train_batch_sampler = make_batch_data_sampler(train_sampler, args.batch_size, args.max_iters)
# self.train_loader = data.DataLoader(dataset=trainset,
# batch_sampler=train_batch_sampler,
# num_workers=args.workers,
# pin_memory=True)
self.train_loader = data.DataLoader(dataset=train_set,
batch_sampler=train_batch_sampler,
#batch_size = args.batch_size,
num_workers=args.workers,
# shuffle = True,
pin_memory=True)
if not args.skip_val:
# valset = get_segmentation_dataset(args.dataset, split='val', mode='val', **data_kwargs)
# valset = get_segmentation_dataset(args.dataset,
# images_dir='/home/wangjialei/teeth_dataset/new_data_20190621/valid_new/images',
# masks_dir='/home/wangjialei/teeth_dataset/new_data_20190621/valid_new/masks',
# nb_classes=3, mode='val', **data_kwargs)
valid_img_dir = '/home/wangjialei/teeth_dataset/new_data_20190621/valid_new/images'
valid_mask_dir = '/home/wangjialei/teeth_dataset/new_data_20190621/valid_new/masks'
valid_transform=transforms.Compose([
transforms.ToTensor(),
transforms.Normalize([0.517446, 0.360147, 0.310427], [0.061526,0.049087, 0.041330])#R_var is 0.061526, G_var is 0.049087, B_var is 0.041330
])
data_kwargs = {'transform': valid_transform, 'base_size': args.base_size, 'crop_size': args.crop_size}
# valid_set = get_segmentation_dataset(args.dataset, valid_img_dir, valid_mask_dir, self.nb_classes, 'valid', **data_kwargs)
valid_set = SegmentationData(images_dir=valid_img_dir, masks_dir=valid_mask_dir, nb_classes=self.nb_classes, mode='valid', transform=valid_transform)
valid_sampler = make_data_sampler(valid_set, False, args.distributed)
valid_batch_sampler = make_batch_data_sampler(valid_sampler, args.batch_size)
# self.val_loader = data.DataLoader(dataset=valset,
# batch_sampler=val_batch_sampler,
# num_workers=args.workers,
# pin_memory=True)
self.valid_loader = data.DataLoader(dataset=valid_set,
batch_sampler=valid_batch_sampler,
# batch_size = args.batch_size,
num_workers=args.workers,
# shuffle = False,
pin_memory=True)
# create network
BatchNorm2d = nn.SyncBatchNorm if args.distributed else nn.BatchNorm2d
self.model = get_segmentation_model(args.model, dataset=args.dataset,
aux=args.aux, norm_layer=BatchNorm2d).to(self.device)
# resume checkpoint if needed
if args.resume:
if os.path.isfile(args.resume):
name, ext = os.path.splitext(args.resume)
assert ext == '.pkl' or '.pth', 'Sorry only .pth and .pkl files supported.'
print('Resuming training, loading {}...'.format(args.resume))
self.model.load_state_dict(torch.load(args.resume, map_location=lambda storage, loc: storage))
# create criterion
if args.ohem:
min_kept = int(args.batch_size // args.num_gpus * args.crop_size ** 2 // 16)
self.criterion = MixSoftmaxCrossEntropyOHEMLoss(args.aux, args.aux_weight, min_kept=min_kept,
ignore_index=-1).to(self.device)
else:
self.criterion = MixSoftmaxCrossEntropyLoss(args.aux, args.aux_weight, ignore_index=-1).to(self.device)
# optimizer
self.optimizer = torch.optim.SGD(self.model.parameters(),
lr=args.lr,
momentum=args.momentum,
weight_decay=args.weight_decay)
# lr scheduling
self.lr_scheduler = WarmupPolyLR(self.optimizer,
max_iters=args.max_iters,
power=0.9,
warmup_factor=args.warmup_factor,
warmup_iters=args.warmup_iters,
warmup_method=args.warmup_method)
if args.distributed:
self.model = nn.parallel.DistributedDataParallel(self.model,
device_ids=[args.local_rank],
output_device=args.local_rank,
find_unused_parameters=True)
# evaluation metrics
self.metric = SegmentationMetric(train_set.num_class)
self.best_pred = 0.0
class Evaluator(object):
def __init__(self, args):
self.args = args
self.device = torch.device(args.device)
# image transform
input_transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize([.485, .456, .406], [.229, .224, .225]),
])
# dataset and dataloader
val_dataset = get_segmentation_dataset('eyes',
split='val',
mode='testval',
transform=input_transform)
val_sampler = make_data_sampler(val_dataset, False, args.distributed)
val_batch_sampler = make_batch_data_sampler(val_sampler,
images_per_batch=1)
self.val_loader = data.DataLoader(dataset=val_dataset,
batch_sampler=val_batch_sampler,
num_workers=args.workers,
pin_memory=True)
# create network
self.model = get_segmentation_model(model=args.model,
dataset=args.dataset,
aux=args.aux,
pretrained=True,
pretrained_base=False)
if args.distributed:
self.model = self.model.module
self.model.to(self.device)
self.metric = SegmentationMetric(val_dataset.num_class)
def eval(self):
self.metric.reset()
self.model.eval()
if self.args.distributed:
model = self.model.module
else:
model = self.model
logger.info("Start validation, Total sample: {:d}".format(
len(self.val_loader)))
for i, (image, target) in enumerate(self.val_loader):
image = image.to(self.device)
target = target.to(self.device)
with torch.no_grad():
outputs = model(image)
self.metric.update(outputs[0], target)
pixAcc, mIoU = self.metric.get()
logger.info(
"Sample: {:d}, validation pixAcc: {:.3f}, mIoU: {:.3f}".format(
i + 1, pixAcc * 100, mIoU * 100))
if True:
pred = torch.argmax(outputs[0], 1)
pred = pred.cpu().data.numpy()
predict = pred.squeeze(0)
# mask = get_color_pallete(predict, self.args.dataset)
image = image.cpu().data.numpy().squeeze(0).transpose(
(1, 2, 0))
image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
colors = np.array([[255, 0, 0], [0, 255, 0], [0, 0, 255]])
res = np.zeros((image.shape[0] * 3, image.shape[1], 3))
inp = ((image + 1) * 127.5).astype('int')
msk = colors[predict]
res[0:image.shape[0], :, :] = inp
res[image.shape[0]:image.shape[0] * 2, :, :] = msk
res[image.shape[0] * 2:, :, :] = cv2.addWeighted(
inp, 0.5, msk, 0.5, 0)
cv2.imwrite(
f'/root/mitya/Lightweight-Segmentation/results/{i}.png',
res)
synchronize()
class Evaluator(object):
def __init__(self, args):
self.args = args
self.device = torch.device(args.device)
# image transform
input_transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize([.485, .456, .406], [.229, .224, .225]),
])
# dataset and dataloader
val60_data_kwargs = {
'transform': input_transform,
'base_size': args.base_size,
'crop_size': args.crop_size,
're_size': args.re_size,
}
valset = get_segmentation_dataset(args.dataset,
args=args,
split='val',
mode='val_onlyrs',
**val60_data_kwargs)
val_sampler = make_data_sampler(valset, True, args.distributed)
val_batch_sampler = make_batch_data_sampler(val_sampler,
args.batch_size)
self.val60_loader = data.DataLoader(dataset=valset,
batch_sampler=val_batch_sampler,
num_workers=args.workers,
pin_memory=True)
# create network
BatchNorm2d = nn.SyncBatchNorm if args.distributed else nn.BatchNorm2d
self.model = get_segmentation_model(args.model,
dataset=args.dataset,
args=self.args,
norm_layer=BatchNorm2d).to(
self.device)
self.model = load_model(args.resume, self.model)
# evaluation metrics
self.metric_120 = SegmentationMetric(valset.num_class)
self.metric_60 = SegmentationMetric(valset.num_class)
self.best_pred = 0.0
def evaluate(self):
is_best = False
self.metric_120.reset()
self.metric_60.reset()
if self.args.distributed:
model = self.model.module
else:
model = self.model
torch.cuda.empty_cache() # TODO check if it helps
model.eval()
loss = [[], []]
for i, (image, target, _) in enumerate(self.val60_loader):
for index in range(len(image)):
image[index] = image[index].to(self.device)
for index in range(len(target)):
target[index] = target[index].to(self.device)
with torch.no_grad():
outputs = model(image)
self.metric_120.update(outputs[0][0], target[0])
self.metric_60.update(outputs[0][1], target[1])
if self.args.save_pre:
self.save_pred(image, target, _, outputs)
pixAcc_120, mIoU_120, Iou_120 = self.metric_120.get()
val_mIou_120 = mIoU_120
val_mpixAcc_120 = pixAcc_120
logger.info("120 Validation: mpixAcc: {:.3f}, mIoU: {:.3f}".format(
val_mpixAcc_120, val_mIou_120))
for i, j in enumerate(Iou_120):
logger.info("class {:d} : {:.3f}".format(i, j))
pixAcc_60, mIoU_60, Iou_60 = self.metric_60.get()
val_mIou_60 = mIoU_60
val_mpixAcc_60 = pixAcc_60
logger.info("60 Validation: mpixAcc: {:.3f}, mIoU: {:.3f}".format(
val_mpixAcc_60, val_mIou_60))
for i, j in enumerate(Iou_60):
logger.info("class {:d} : {:.3f}".format(i, j))
synchronize()
def save_pred(self, image, target, image_name, outputs):
def unnormlize(img, mean, std):
mean = np.expand_dims(mean, axis=0)
mean = np.repeat(mean, img.shape[1], axis=0)
mean = np.expand_dims(mean, axis=0)
mean = np.repeat(mean, img.shape[0], axis=0)
std = np.expand_dims(std, axis=0)
std = np.repeat(std, img.shape[1], axis=0)
std = np.expand_dims(std, axis=0)
std = np.repeat(std, img.shape[0], axis=0)
img = (img * std + mean) * 255.
return img
mean = np.array([.485, .456, .406])
std = np.array([.229, .224, .225])
################################### 120 ##########################################
pred_120 = torch.argmax(outputs[0][0], 1)
pred_120 = pred_120.cpu().data.numpy()
predict_120 = pred_120.squeeze(0)
mask_120 = get_color_pallete(predict_120, self.args.dataset)
mask_120 = np.asarray(mask_120.convert('RGB'))
misc.imsave(
os.path.join(
self.args.save_pre_path,
str(image_name[1])[2:-2] + '_' + self.args.model_mode +
'.png'), mask_120)
if self.args.combined:
image_120 = image[0]
image_120 = image_120.cpu().data.numpy()[0].transpose(1, 2, 0)
image_120 = np.array(unnormlize(image_120, mean, std),
dtype=np.int32)
target_120 = target[0].cpu().data.numpy()
target_120 = target_120.squeeze(0)
target_120 = get_color_pallete(target_120, self.args.dataset)
target_120 = np.asarray(target_120.convert('RGB'))
combine1 = np.concatenate(
(image_120, image_120 * 0.5 + mask_120 * 0.5), axis=1)
combine2 = np.concatenate((target_120, mask_120), axis=1)
mask_120 = np.concatenate((combine1, combine2), axis=0)
misc.imsave(
os.path.join(
self.args.save_pre_path,
str(image_name[0])[2:-2] + '_' + self.args.model_mode +
'_4.png'), mask_120)
################################### 60 ##########################################
pred_60 = torch.argmax(outputs[0][1], 1)
pred_60 = pred_60.cpu().data.numpy()
predict_60 = pred_60.squeeze(0)
mask_60 = get_color_pallete(predict_60, self.args.dataset)
mask_60 = np.asarray(mask_60.convert('RGB'))
misc.imsave(
os.path.join(
self.args.save_pre_path,
str(image_name[1])[2:-2] + '_' + self.args.model_mode +
'.png'), mask_60)
if self.args.combined:
image_60 = image[1]
image_60 = image_60.cpu().data.numpy()[0].transpose(1, 2, 0)
image_60 = np.array(unnormlize(image_60, mean, std),
dtype=np.int32)
target_60 = target[1].cpu().data.numpy()
target_60 = target_60.squeeze(0)
target_60 = get_color_pallete(target_60, self.args.dataset)
target_60 = np.asarray(target_60.convert('RGB'))
combine1 = np.concatenate(
(image_60, image_60 * 0.5 + mask_60 * 0.5), axis=1)
combine2 = np.concatenate((target_60, mask_60), axis=1)
mask_60 = np.concatenate((combine1, combine2), axis=0)
misc.imsave(
os.path.join(
self.args.save_pre_path,
str(image_name[1])[2:-2] + '_' + self.args.model_mode +
'_4.png'), mask_60)
class Evaluator(object):
def __init__(self, args):
self.args = args
self.device = torch.device(args.device)
self.nb_classes = 3
valid_img_dir = '/home/wangjialei/teeth_dataset/new_data_20190621/valid_new/images'
valid_mask_dir = '/home/wangjialei/teeth_dataset/new_data_20190621/valid_new/masks'
# valid_transform=transforms.Compose([
# # transforms.ToTensor(),
# # transforms.Normalize([0.517446, 0.360147, 0.310427], [0.061526,0.049087, 0.041330])#R_var is 0.061526, G_var is 0.049087, B_var is 0.041330
# ])
# dataset and dataloader
valid_set = SegmentationData(images_dir=valid_img_dir,
masks_dir=valid_mask_dir,
nb_classes=self.nb_classes,
mode='valid',
transform=None)
valid_sampler = make_data_sampler(valid_set, False, args.distributed)
valid_batch_sampler = make_batch_data_sampler(valid_sampler,
images_per_batch=1)
self.val_loader = data.DataLoader(dataset=valid_set,
batch_sampler=valid_batch_sampler,
num_workers=args.workers,
pin_memory=True)
# create network
self.model = get_segmentation_model(model=args.model,
dataset=args.dataset,
aux=args.aux,
pretrained=True,
pretrained_base=False)
if args.distributed:
self.model = self.model.module
self.model.to(self.device)
self.metric = SegmentationMetric(valid_set.num_class)
def eval(self):
self.metric.reset()
self.model.eval()
if self.args.distributed:
model = self.model.module
else:
model = self.model
logger.info("Start validation, Total sample: {:d}".format(
len(self.val_loader)))
for i, (image, target) in enumerate(self.val_loader):
img = data_process(image)
img = img.to(self.device)
target = target.to(self.device)
with torch.no_grad():
outputs = model(img)
self.metric.update(outputs, target)
pixAcc, mIoU = self.metric.get()
logger.info(
"Sample: {:d}, validation pixAcc: {:.3f}, mIoU: {:.3f}".format(
i + 1, pixAcc * 100, mIoU * 100))
if self.args.save_pred:
pred = torch.argmax(outputs[0], 1)
pred = pred.cpu().data.numpy()
predict = pred.squeeze(0)
img_show = image[0].numpy()
img_show = img_show.astype('uint8')
plt.subplot(1, 3, 1)
plt.title('image')
plt.imshow(img_show)
mask = target.cpu().data.numpy()
mask = mask.reshape(mask.shape[1], mask.shape[2])
mask = mask_to_image(mask)
plt.subplot(1, 3, 2)
plt.title('mask')
plt.imshow(mask)
predict = mask_to_image(predict)
plt.subplot(1, 3, 3)
plt.title('pred')
plt.imshow(predict)
save_file = "save_fig_val"
os.makedirs(save_file, exist_ok=True)
plt.savefig(os.path.join(save_file, str(i) + '.png'))
synchronize()
def test(self):
self.model.eval()
if self.args.distributed:
model = self.model.module
else:
model = self.model
test_img_dir = '/home/wangjialei/projects/teeth_bad_case/'
img_folder = os.listdir(test_img_dir)
for iter, img_file in enumerate(img_folder):
img_name = test_img_dir + img_file
image = Image.open(img_name)
print(type(image))
img = data_process(image)
img = img.to(self.device)
with torch.no_grad():
outputs = model(img)
if self.args.save_pred:
pred = torch.argmax(outputs[0], 1)
pred = pred.cpu().data.numpy()
predict = pred.squeeze(0)
img_show = image
plt.subplot(1, 2, 1)
plt.title('image')
plt.imshow(img_show)
predict = mask_to_image(predict)
plt.subplot(1, 2, 2)
plt.title('pred')
plt.imshow(predict)
save_file = "save_fig_test"
os.makedirs(save_file, exist_ok=True)
plt.savefig(os.path.join(save_file, str(iter) + '.png'))
class Evaluator(object):
def __init__(self, args):
self.args = args
self.device = torch.device(args.device)
# image transform
input_transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize([.485, .456, .406], [.229, .224, .225]),
])
# dataset and dataloader
val_dataset = get_segmentation_dataset(args.dataset,
split='val',
mode='testval',
transform=input_transform)
val_sampler = make_data_sampler(val_dataset, False, args.distributed)
val_batch_sampler = make_batch_data_sampler(val_sampler,
images_per_batch=1)
self.val_loader = data.DataLoader(dataset=val_dataset,
batch_sampler=val_batch_sampler,
num_workers=args.workers,
pin_memory=True)
# create network
self.model = get_segmentation_model(model=args.model,
dataset=args.dataset,
aux=args.aux,
pretrained=True,
pretrained_base=False)
if args.distributed:
self.model = self.model.module
self.model.to(self.device)
self.metric = SegmentationMetric(val_dataset.num_class)
def eval(self):
self.metric.reset()
self.model.eval()
if self.args.distributed:
model = self.model.module
else:
model = self.model
logger.info("Start validation, Total sample: {:d}".format(
len(self.val_loader)))
for i, (image, target) in enumerate(self.val_loader):
image = image.to(self.device)
target = target.to(self.device)
with torch.no_grad():
outputs = model(image)
self.metric.update(outputs[0], target)
pixAcc, mIoU = self.metric.get()
logger.info(
"Sample: {:d}, validation pixAcc: {:.3f}, mIoU: {:.3f}".format(
i + 1, pixAcc * 100, mIoU * 100))
if self.args.save_pred:
pred = torch.argmax(outputs[0], 1)
pred = pred.cpu().data.numpy()
predict = pred.squeeze(0)
mask = get_color_pallete(predict, self.args.dataset)
# mask.save(os.path.join(outdir, os.path.splitext(filename[0])[0] + '.png'))
synchronize()
