class Metrics(object):
def __int__(self):
self.nme = None
self.auc = None
self.loss = None
@metric
def add_nme(self, decay=0.999):
self.nme = NME(decay)
@metric
def add_auc(self, low=0, high=0.08, step=0.01, decay=0.99):
self.auc = AUC(low, high, step, decay)
@metric
def add_loss(self, decay=0.999):
self.loss = Loss(decay)
def clear(self):
self.nme.clear()
self.auc.clear()
self.loss.clear()
def main(data_path, batch_size, lr, n_epoch, kernel_size, filter_start,
sh_degree, depth, n_side, rf_name, wm, gm, csf, loss_fn_intensity,
loss_fn_non_negativity, loss_fn_sparsity, sigma_sparsity,
intensity_weight, nn_fodf_weight, sparsity_weight, save_path,
save_every, normalize, load_state):
"""Train a model
Args:
data_path (str): Data path
batch_size (int): Batch size
lr (float): Learning rate
n_epoch (int): Number of training epoch
kernel_size (int): Kernel Size
filter_start (int): Number of output features of the first convolution layer
sh_degree (int): Spherical harmonic degree of the fODF
depth (int): Graph subsample depth
n_side (int): Resolution of the Healpix map
rf_name (str): Response function algorithm name
wm (float): Use white matter
gm (float): Use gray matter
csf (float): Use CSF
loss_fn_intensity (str): Name of the intensity loss
loss_fn_non_negativity (str): Name of the nn loss
loss_fn_sparsity (str): Name of the sparsity loss
intensity_weight (float): Weight of the intensity loss
nn_fodf_weight (float): Weight of the nn loss
sparsity_weight (float): Weight of the sparsity loss
save_path (str): Save path
save_every (int): Frequency to save the model
normalize (bool): Normalize the fODFs
load_state (str): Load pre trained network
"""
# Load the shell and the graph samplings
shellSampling = ShellSampling(f'{data_path}/bvecs.bvecs',
f'{data_path}/bvals.bvals',
sh_degree=sh_degree,
max_sh_degree=8)
graphSampling = HealpixSampling(n_side, depth, sh_degree=sh_degree)
# Load the image and the mask
dataset = DMRIDataset(f'{data_path}/features.nii', f'{data_path}/mask.nii')
dataloader_train = DataLoader(dataset=dataset,
batch_size=batch_size,
shuffle=True)
n_batch = len(dataloader_train)
# Load the Polar filter used for the deconvolution
polar_filter_equi, polar_filter_inva = load_response_function(
f'{data_path}/response_functions/{rf_name}',
wm=wm,
gm=gm,
csf=csf,
max_degree=sh_degree,
n_shell=len(shellSampling.shell_values))
# Create the deconvolution model
model = Model(polar_filter_equi, polar_filter_inva, shellSampling,
graphSampling, filter_start, kernel_size, normalize)
if load_state:
print(load_state)
model.load_state_dict(torch.load(load_state), strict=False)
# Load model in GPU
model = model.to(DEVICE)
torch.save(model.state_dict(),
os.path.join(save_path, 'history', 'epoch_0.pth'))
# Loss
intensity_criterion = Loss(loss_fn_intensity)
non_negativity_criterion = Loss(loss_fn_non_negativity)
sparsity_criterion = Loss(loss_fn_sparsity, sigma_sparsity)
# Create dense interpolation used for the non-negativity and the sparsity losses
denseGrid_interpolate = ComputeSignal(
torch.Tensor(graphSampling.sampling.SH2S))
denseGrid_interpolate = denseGrid_interpolate.to(DEVICE)
# Optimizer/Scheduler
optimizer = torch.optim.Adam(model.parameters(), lr=lr)
scheduler = ReduceLROnPlateau(optimizer,
threshold=0.01,
factor=0.1,
patience=3,
verbose=True)
save_loss = {}
save_loss['train'] = {}
writer = SummaryWriter(log_dir=os.path.join(data_path, 'result', 'run',
save_path.split('/')[-1]))
tb_j = 0
# Training loop
for epoch in range(n_epoch):
# TRAIN
model.train()
# Initialize loss to save and plot.
loss_intensity_ = 0
loss_sparsity_ = 0
loss_non_negativity_fodf_ = 0
# Train on batch.
for batch, data in enumerate(dataloader_train):
# Delete all previous gradients
optimizer.zero_grad()
to_print = ''
# Load the data in the DEVICE
input = data['input'].to(DEVICE)
output = data['output'].to(DEVICE)
mask = data['mask'].to(DEVICE)
x_reconstructed, x_deconvolved_equi_shc, x_deconvolved_inva_shc = model(
input)
###############################################################################################
###############################################################################################
# Loss
###############################################################################################
###############################################################################################
# Intensity loss
loss_intensity = intensity_criterion(x_reconstructed, output, mask)
loss_intensity_ += loss_intensity.item()
loss = intensity_weight * loss_intensity
to_print += ', Intensity: {0:.10f}'.format(loss_intensity.item())
if not x_deconvolved_equi_shc is None:
x_deconvolved_equi = denseGrid_interpolate(
x_deconvolved_equi_shc)
###############################################################################################
# Sparsity loss
equi_sparse = torch.zeros(x_deconvolved_equi.shape).to(DEVICE)
loss_sparsity = sparsity_criterion(x_deconvolved_equi,
equi_sparse, mask)
loss_sparsity_ += loss_sparsity.item()
loss += sparsity_weight * loss_sparsity
to_print += ', Equi Sparsity: {0:.10f}'.format(
loss_sparsity.item())
###############################################################################################
# Non negativity loss
fodf_neg = torch.min(x_deconvolved_equi,
torch.zeros_like(x_deconvolved_equi))
fodf_neg_zeros = torch.zeros(fodf_neg.shape).to(DEVICE)
loss_non_negativity_fodf = non_negativity_criterion(
fodf_neg, fodf_neg_zeros, mask)
loss_non_negativity_fodf_ += loss_non_negativity_fodf.item()
loss += nn_fodf_weight * loss_non_negativity_fodf
to_print += ', Equi NN: {0:.10f}'.format(
loss_non_negativity_fodf.item())
###############################################################################################
# Partial volume regularizer
regularizer_equi = 0.00001 * 1 / torch.mean(
x_deconvolved_equi_shc[mask == 1][:, :, 0]) * np.sqrt(
4 * np.pi)
loss += regularizer_equi
to_print += ', Equi regularizer: {0:.10f}'.format(
regularizer_equi.item())
if not x_deconvolved_inva_shc is None:
###############################################################################################
# Partial volume regularizer
regularizer_inva = 0.00001 * 1 / torch.mean(
x_deconvolved_inva_shc[mask == 1][:, :, 0]) * np.sqrt(
4 * np.pi)
loss += regularizer_inva
to_print += ', Inva regularizer: {0:.10f}'.format(
regularizer_inva.item())
###############################################################################################
# Tensorboard
tb_j += 1
writer.add_scalar('Batch/train_intensity', loss_intensity.item(),
tb_j)
writer.add_scalar('Batch/train_sparsity', loss_sparsity.item(),
tb_j)
writer.add_scalar('Batch/train_nn',
loss_non_negativity_fodf.item(), tb_j)
writer.add_scalar('Batch/train_total', loss.item(), tb_j)
###############################################################################################
# To print loss
to_print = 'Epoch [{0}/{1}], Iter [{2}/{3}]: Loss: {4:.10f}'.format(epoch + 1, n_epoch,
batch + 1, n_batch,
loss.item()) \
+ to_print
print(to_print, end="\r")
###############################################################################################
# Loss backward
loss = loss
loss.backward()
optimizer.step()
if (batch + 1) % 500 == 0:
torch.save(
model.state_dict(),
os.path.join(save_path, 'history',
'epoch_{0}.pth'.format(epoch + 1)))
###############################################################################################
# Save and print mean loss for the epoch
print("")
to_print = ''
loss_ = 0
# Mean results of the last epoch
save_loss['train'][epoch] = {}
save_loss['train'][epoch]['loss_intensity'] = loss_intensity_ / n_batch
save_loss['train'][epoch]['weight_loss_intensity'] = intensity_weight
loss_ += intensity_weight * loss_intensity_
to_print += ', Intensity: {0:.10f}'.format(loss_intensity_ / n_batch)
save_loss['train'][epoch]['loss_sparsity'] = loss_sparsity_ / n_batch
save_loss['train'][epoch]['weight_loss_sparsity'] = sparsity_weight
loss_ += sparsity_weight * loss_sparsity_
to_print += ', Sparsity: {0:.10f}'.format(loss_sparsity_ / n_batch)
save_loss['train'][epoch][
'loss_non_negativity_fodf'] = loss_non_negativity_fodf_ / n_batch
save_loss['train'][epoch][
'weight_loss_non_negativity_fodf'] = nn_fodf_weight
loss_ += nn_fodf_weight * loss_non_negativity_fodf_
to_print += ', WM fODF NN: {0:.10f}'.format(loss_non_negativity_fodf_ /
n_batch)
save_loss['train'][epoch]['loss'] = loss_ / n_batch
to_print = 'Epoch [{0}/{1}], Train Loss: {2:.10f}'.format(
epoch + 1, n_epoch, loss_ / n_batch) + to_print
print(to_print)
writer.add_scalar('Epoch/train_intensity', loss_intensity_ / n_batch,
epoch)
writer.add_scalar('Epoch/train_sparsity', loss_sparsity_ / n_batch,
epoch)
writer.add_scalar('Epoch/train_nn',
loss_non_negativity_fodf_ / n_batch, epoch)
writer.add_scalar('Epoch/train_total', loss_ / n_batch, epoch)
###############################################################################################
# VALIDATION
scheduler.step(loss_ / n_batch)
if epoch == 0:
min_loss = loss_
epochs_no_improve = 0
n_epochs_stop = 5
early_stop = False
elif loss_ < min_loss * 0.999:
epochs_no_improve = 0
min_loss = loss_
else:
epochs_no_improve += 1
if epoch > 5 and epochs_no_improve == n_epochs_stop:
print('Early stopping!')
early_stop = True
###############################################################################################
# Save the loss and model
with open(os.path.join(save_path, 'history', 'loss.pkl'), 'wb') as f:
pickle.dump(save_loss, f)
if (epoch + 1) % save_every == 0:
torch.save(
model.state_dict(),
os.path.join(save_path, 'history',
'epoch_{0}.pth'.format(epoch + 1)))
if early_stop:
print("Stopped")
break
def train(train_img_path, train_gt_path, pths_path, batch_size, lr,
num_workers, epoch_iter, interval, checkpoint, eval_interval,
test_img_path, submit_path):
file_num = len(os.listdir(train_img_path))
trainset = custom_dataset(train_img_path, train_gt_path)
train_loader = data.DataLoader(trainset, batch_size=batch_size, \
shuffle = True, num_workers=num_workers, drop_last=True)
criterion = Loss()
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
model = EAST(pretrained=False)
if checkpoint:
model.load_state_dict(torch.load(checkpoint))
data_parallel = False
if torch.cuda.device_count() > 1:
model = nn.DataParallel(model)
# model = DataParallelModel(model)
data_parallel = True
model.to(device)
optimizer = torch.optim.Adam(model.parameters(), lr=lr)
# optimizer = torch.optim.SGD(model.parameters(), lr=lr, momentum=0.9,weight_decay=0)
scheduler = lr_scheduler.MultiStepLR(optimizer,
milestones=[epoch_iter // 2],
gamma=0.1)
whole_number = epoch_iter * (len(trainset) / batch_size)
print("epoch size:%d" % (epoch_iter))
print("batch size:%d" % (batch_size))
print("data number:%d" % (len(trainset)))
all_loss = []
current_i = 0
for epoch in range(epoch_iter):
model.train()
epoch_loss = 0
epoch_time = time.time()
for i, (img, gt_score, gt_geo, ignored_map,
_) in enumerate(train_loader):
current_i += 1
start_time = time.time()
img, gt_score, gt_geo, ignored_map = img.to(device), gt_score.to(
device), gt_geo.to(device), ignored_map.to(device)
pred_score, pred_geo = model(img)
loss = criterion(gt_score, pred_score, gt_geo, pred_geo,
ignored_map)
epoch_loss += loss.item()
optimizer.zero_grad()
loss.backward()
optimizer.step()
lr_now = scheduler.get_last_lr()
progress_bar(40, loss.item(), current_i, whole_number, lr_now[0])
scheduler.step()
print('epoch_loss is {:.8f}, epoch_time is {:.8f}'.format(
epoch_loss / int(file_num / batch_size),
time.time() - epoch_time))
all_loss.append(epoch_loss / int(file_num / batch_size))
print(time.asctime(time.localtime(time.time())))
plt.plot(all_loss)
plt.savefig('loss_landscape.png')
plt.close()
print('=' * 50)
if (epoch + 1) % interval == 0:
state_dict = model.module.state_dict(
) if data_parallel else model.state_dict()
torch.save(
state_dict,
os.path.join(pths_path,
'model_epoch_{}.pth'.format(epoch + 1)))
output = open(os.path.join(pths_path, 'loss.pkl'), 'wb')
pkl.dump(all_loss, output)
def __init__(self, batch_size=32, optimizer_name="Adam", lr=1e-3, weight_decay=1e-5,
epochs=200, model_name="model01", gpu_ids=None, resume=None, tqdm=None):
"""
args:
batch_size = (int) batch_size of training and validation
lr = (float) learning rate of optimization
weight_decay = (float) weight decay of optimization
epochs = (int) The number of epochs of training
model_name = (string) The name of training model. Will be folder name.
gpu_ids = (List) List of gpu_ids. (e.g. gpu_ids = [0, 1]). Use CPU, if it is None.
resume = (Dict) Dict of some settings. (resume = {"checkpoint_path":PATH_of_checkpoint, "fine_tuning":True or False}).
Learn from scratch, if it is None.
tqdm = (tqdm Object) progress bar object. Set your tqdm please.
Don't view progress bar, if it is None.
"""
# Set params
self.batch_size = batch_size
self.epochs = epochs
self.start_epoch = 0
self.use_cuda = (gpu_ids is not None) and torch.cuda.is_available
self.tqdm = tqdm
self.use_tqdm = tqdm is not None
# ------------------------- #
# Define Utils. (No need to Change.)
"""
These are Project Modules.
You may not have to change these.
Saver: Save model weight. / <utils.saver.Saver()>
TensorboardSummary: Write tensorboard file. / <utils.summaries.TensorboardSummary()>
Evaluator: Calculate some metrics (e.g. Accuracy). / <utils.metrics.Evaluator()>
"""
## ***Define Saver***
self.saver = Saver(model_name, lr, epochs)
self.saver.save_experiment_config()
## ***Define Tensorboard Summary***
self.summary = TensorboardSummary(self.saver.experiment_dir)
self.writer = self.summary.create_summary()
# ------------------------- #
# Define Training components. (You have to Change!)
"""
These are important setting for training.
You have to change these.
make_data_loader: This creates some <Dataloader>s. / <dataloader.__init__>
Modeling: You have to define your Model. / <modeling.modeling.Modeling()>
Evaluator: You have to define Evaluator. / <utils.metrics.Evaluator()>
Optimizer: You have to define Optimizer. / <utils.optimizer.Optimizer()>
Loss: You have to define Loss function. / <utils.loss.Loss()>
"""
## ***Define Dataloader***
self.train_loader, self.val_loader, self.test_loader, self.num_classes = make_data_loader(batch_size)
## ***Define Your Model***
self.model = Modeling(self.num_classes)
## ***Define Evaluator***
self.evaluator = Evaluator(self.num_classes)
## ***Define Optimizer***
self.optimizer = Optimizer(self.model.parameters(), optimizer_name=optimizer_name, lr=lr, weight_decay=weight_decay)
## ***Define Loss***
self.criterion = Loss()
# ------------------------- #
# Some settings
"""
You don't have to touch bellow code.
Using cuda: Enable to use cuda if you want.
Resuming checkpoint: You can resume training if you want.
"""
## ***Using cuda***
if self.use_cuda:
self.model = torch.nn.DataParallel(self.model, device_ids=gpu_ids).cuda()
## ***Resuming checkpoint***
"""You can ignore bellow code."""
self.best_pred = 0.0
if resume is not None:
if not os.path.isfile(resume["checkpoint_path"]):
raise RuntimeError("=> no checkpoint found at '{}'" .format(resume["checkpoint_path"]))
checkpoint = torch.load(resume["checkpoint_path"])
self.start_epoch = checkpoint['epoch']
if self.use_cuda:
self.model.module.load_state_dict(checkpoint['state_dict'])
else:
self.model.load_state_dict(checkpoint['state_dict'])
if resume["fine_tuning"]:
# resume params of optimizer, if run fine tuning.
self.optimizer.load_state_dict(checkpoint['optimizer'])
self.start_epoch = 0
self.best_pred = checkpoint['best_pred']
print("=> loaded checkpoint '{}' (epoch {})"
.format(resume["checkpoint_path"], checkpoint['epoch']))
from utils.loss import Loss
from utils.dataset import Yolo_dataset
from config import cfg
from utils.test_mAP import evaluate
if __name__ == "__main__":
scaler = torch.cuda.amp.GradScaler()
torch.backends.cudnn.benchmark = True
anchors = np.array(cfg.anchors).reshape([-1, 2])
model = YOLOv4(cfg.cfgfile).to(cfg.device)
# model.load_weights('yolov4.weights')
# model.load_state_dict(torch.load("weights/Epoch50-Total_Loss2.2437.pth"))
vis = visdom.Visdom(env='YOLOv4')
yolo_losses = []
for i in range(3):
yolo_losses.append(Loss(i))
optimizer = optim.Adam(model.parameters(), cfg.lr)
if cfg.Cosine_lr:
lr_scheduler = optim.lr_scheduler.CosineAnnealingLR(optimizer,
T_max=5,
eta_min=1e-5)
else:
lr_scheduler = optim.lr_scheduler.StepLR(optimizer,
step_size=1,
gamma=0.96)
train_dataset = Yolo_dataset(train=True)
val_dataset = Yolo_dataset(train=False)
train_loader = DataLoader(train_dataset,
shuffle=True,
def main():
global args
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID" # see issue #152
os.environ["CUDA_VISIBLE_DEVICES"] = args.gpu
__normalize = {'mean': (0.485, 0.456, 0.406), 'std': (0.229, 0.224, 0.225)}
TrainImgLoader = torch.utils.data.DataLoader(DA(args.datapath,
split='train',
normalize=__normalize),
batch_size=args.train_bsize,
shuffle=True,
num_workers=1,
drop_last=False)
ValImgLoader = torch.utils.data.DataLoader(DA(args.datapath,
split='val',
normalize=__normalize),
batch_size=args.test_bsize,
shuffle=False,
num_workers=1,
drop_last=False)
TestImgLoader = torch.utils.data.DataLoader(DA(args.datapath,
split='test',
normalize=__normalize),
batch_size=args.test_bsize,
shuffle=False,
num_workers=1,
drop_last=False)
if not os.path.isdir(args.save_path):
os.makedirs(os.path.join(args.save_path, 'train'))
os.makedirs(os.path.join(args.save_path, 'test'))
os.makedirs(os.path.join(args.save_path, 'val'))
log = logger.setup_logger(args.save_path + '/training.log')
writer = logger.setup_tensorboard(args.save_path)
for key, value in sorted(vars(args).items()):
log.info(str(key) + ':' + str(value))
model = StereoNet(k=args.stages - 1,
r=args.stages - 1,
maxdisp=args.maxdisp)
model = nn.DataParallel(model).cuda()
model.apply(weights_init)
criterion = Loss(args)
optimizer = optim.RMSprop(model.parameters(), lr=args.lr)
scheduler = lr_scheduler.StepLR(optimizer,
step_size=args.stepsize,
gamma=args.gamma)
log.info('Number of model parameters: {}'.format(
sum([p.data.nelement() for p in model.parameters()])))
args.start_epoch = 0
if args.resume:
if os.path.isfile(args.resume):
log.info("=> loading checkpoint '{}'".format((args.resume)))
checkpoint = torch.load(args.resume)
args.start_epoch = checkpoint['epoch']
model.load_state_dict(checkpoint['state_dict'])
optimizer.load_state_dict(checkpoint['optimizer'])
log.info("=> loaded checkpoint '{}' (epoch {})".format(
args.resume, checkpoint['epoch']))
else:
log.info("=> no checkpoint found at '{}'".format(args.resume))
log.info("=> will start from scratch.")
else:
log.info("Not Resume")
start_full_time = time.time()
for epoch in range(args.start_epoch, args.epoch):
log.info('This is {}-th epoch'.format(epoch))
train(TrainImgLoader, model, criterion, optimizer, log, writer, epoch)
test(ValImgLoader, model, log, writer, 'val', epoch)
savefilename = args.save_path + '/checkpoint.pth'
torch.save(
{
'epoch': epoch,
'state_dict': model.state_dict(),
'optimizer': optimizer.state_dict()
}, savefilename)
scheduler.step() # will adjust learning rate
test(TestImgLoader, model, log, writer, 'test', epoch)
log.info('full training time = {: 2f} Hours'.format(
(time.time() - start_full_time) / 3600))
schp.save_schp_checkpoint(
{
'state_dict': schp_model.state_dict(),
'cycle_n': cycle_n,
},
False,
"checkpoints",
filename=
f'schp_{opts.model}_{opts.dataset}_cycle{cycle_n}_checkpoint.pth'
)
# schp.save_schp_checkpoint({
# 'state_dict': schp_model.state_dict(),
# 'cycle_n': cycle_n,
# }, False, '/content/drive/MyDrive/', filename=f'schp_{opts.model}_{opts.dataset}_checkpoint.pth')
torch.cuda.empty_cache()
criterion.end_log(len(train_loader))
if __name__ == '__main__':
opts = get_argparser().parse_args(args=[])
if 'ACE2P' in opts.model:
opts.loss_type = 'SCP'
opts.use_mixup = False
os.environ['CUDA_VISIBLE_DEVICES'] = opts.gpu_ids
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
opts.device = device
print("Device: %s" % device)
criterion = Loss(opts)
main(criterion)
criterion.plot_loss('/content/drive/MyDrive/', len(criterion.log))
def add_loss(self, decay=0.999):
self.loss = Loss(decay)
def main(args, logger, summary):
cudnn.enabled = True # Enables bencnmark mode in cudnn, to enable the inbuilt
cudnn.benchmark = True # cudnn auto-tuner to find the best algorithm to use for
# our hardware
seed = random.randint(1, 10000)
logger.info('======>random seed {}'.format(seed))
random.seed(seed) # python random seed
np.random.seed(seed) # set numpy random seed
torch.manual_seed(seed) # set random seed for cpu
# train_set = VaiHinGen(root=args.root, split='trainl',outer_size=2*args.image_size,centre_size=args.image_size)
# test_set = VaiHinGen(root=args.root, split='testl',outer_size=2*args.image_size,centre_size=args.image_size)
train_set = SkmtDataSet(args, split='train')
val_set = SkmtDataSet(args, split='val')
kwargs = {'num_workers': args.workers, 'pin_memory': True}
train_loader = DataLoader(train_set,
batch_size=args.batch_size,
drop_last=True,
shuffle=False,
**kwargs)
test_loader = DataLoader(val_set,
batch_size=1,
drop_last=True,
shuffle=False,
**kwargs)
logger.info('======> building network')
# set model
model = build_skmtnet(backbone='resnet50',
auxiliary_head=args.auxiliary,
trunk_head='deeplab',
num_classes=args.num_classes,
output_stride=16)
logger.info("======> computing network parameters")
total_paramters = netParams(model)
logger.info("the number of parameters: " + str(total_paramters))
# setup optimizer
optimizer = optim.SGD(model.parameters(),
lr=args.lr,
momentum=0.9,
weight_decay=args.weight_decay)
# setup savedir
args.savedir = (args.savedir + '/' + args.model + 'bs' +
str(args.batch_size) + 'gpu' + str(args.gpus) + '/')
if not os.path.exists(args.savedir):
os.makedirs(args.savedir)
# setup optimization criterion
criterion = Loss(args)
if torch.cuda.is_available():
torch.cuda.manual_seed_all(seed) # set random seed for all GPU
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpus
model = nn.DataParallel(model).cuda()
criterion = criterion.cuda()
start_epoch = 0
best_epoch = 0.
best_overall = 0.
best_mIoU = 0.
best_F1 = 0.
trainer = Trainer(args=args,
dataloader=train_loader,
model=model,
optimizer=optimizer,
criterion=criterion,
logger=logger,
summary=summary)
tester = Tester(args=args,
dataloader=test_loader,
model=model,
criterion=criterion,
logger=logger,
summary=summary)
writer = summary.create_summary()
for epoch in range(start_epoch, args.max_epochs):
trainer.train_one_epoch(epoch, writer)
if (epoch % args.show_val_interval == 0):
score, class_iou, class_acc, class_F1 = tester.test_one_epoch(
epoch, writer)
logger.info('======>Now print overall info:')
for k, v in score.items():
logger.info('======>{0:^18} {1:^10}'.format(k, v))
logger.info('======>Now print class acc')
for k, v in class_acc.items():
print('{}: {:.5f}'.format(k, v))
logger.info('======>{0:^18} {1:^10}'.format(k, v))
logger.info('======>Now print class iou')
for k, v in class_iou.items():
print('{}: {:.5f}'.format(k, v))
logger.info('======>{0:^18} {1:^10}'.format(k, v))
logger.info('======>Now print class_F1')
for k, v in class_F1.items():
logger.info('======>{0:^18} {1:^10}'.format(k, v))
if score["Mean IoU(8) : \t"] > best_mIoU:
best_mIoU = score["Mean IoU(8) : \t"]
if score["Overall Acc : \t"] > best_overall:
best_overall = score["Overall Acc : \t"]
# save model in best overall Acc
model_file_name = args.savedir + '/best_model.pth'
torch.save(model.state_dict(), model_file_name)
best_epoch = epoch
if score["Mean F1 : \t"] > best_F1:
best_F1 = score["Mean F1 : \t"]
logger.info("======>best mean IoU:{}".format(best_mIoU))
logger.info("======>best overall : {}".format(best_overall))
logger.info("======>best F1: {}".format(best_F1))
logger.info("======>best epoch: {}".format(best_epoch))
# save the model
model_file_name = args.savedir + '/model.pth'
state = {"epoch": epoch + 1, "model": model.state_dict()}
logger.info('======> Now begining to save model.')
torch.save(state, model_file_name)
logger.info('======> Save done.')