def build_loss(args, datas, ignore_label):
if args.dataset == 'cityscapes':
weight = torch.FloatTensor([0.8373, 0.918, 0.866, 1.0345, 1.0166, 0.9969, 0.9754, 1.0489,
0.8786, 1.0023, 0.9539, 0.9843, 1.1116, 0.9037, 1.0865, 1.0955,
1.0865, 1.1529, 1.0507])
elif datas != None:
weight = torch.from_numpy(datas['classWeights'])
else:
weight = None
# 默认使用交叉商损失函数
criteria = CrossEntropyLoss2d(weight=weight, ignore_label=ignore_label)
if args.loss == 'ProbOhemCrossEntropy2d':
h, w = args.base_size, args.base_size
min_kept = int(args.batch_size // len(args.gpus_id) * h * w // 16)
criteria = ProbOhemCrossEntropy2d(weight=weight, ignore_label=ignore_label, thresh=0.7, min_kept=min_kept)
elif args.loss == 'CrossEntropyLoss2dLabelSmooth':
criteria = CrossEntropyLoss2dLabelSmooth(weight=weight, ignore_label=ignore_label)
# criteria = LabelSmoothing()
elif args.loss == 'LovaszSoftmax':
criteria = LovaszSoftmax(ignore_index=ignore_label)
elif args.loss == 'FocalLoss2d':
criteria = FocalLoss2d(weight=weight, ignore_index=ignore_label)
return criteria
def train_model(args):
"""
args:
args: global arguments
"""
h, w = map(int, args.input_size.split(','))
input_size = (h, w)
print("=====> input size:{}".format(input_size))
print(args)
if args.cuda:
print("=====> use gpu id: '{}'".format(args.gpus))
os.environ["CUDA_VISIBLE_DEVICES"] = args.gpus
if not torch.cuda.is_available():
raise Exception(
"No GPU found or Wrong gpu id, please run without --cuda")
# set the seed
setup_seed(GLOBAL_SEED)
print("=====> set Global Seed: ", GLOBAL_SEED)
cudnn.enabled = True
print("=====> building network")
# build the model and initialization
model = build_model(args.model, num_classes=args.classes)
init_weight(model,
nn.init.kaiming_normal_,
nn.BatchNorm2d,
1e-3,
0.1,
mode='fan_in')
print("=====> computing network parameters and FLOPs")
total_parameters = netParams(model)
print("the number of parameters: %d ==> %.2f M" %
(total_parameters, (total_parameters / 1e6)))
# load data and data augmentation
datas, trainLoader, valLoader = build_dataset_train(
args.dataset, args.classes, input_size, args.batch_size,
args.train_type, False, False, args.num_workers)
args.per_iter = len(trainLoader)
args.max_iter = args.max_epochs * args.per_iter
print('=====> Dataset statistics')
print("data['classWeights']: ", datas['classWeights'])
print('mean and std: ', datas['mean'], datas['std'])
# datas['classWeights'] = np.array([4.044603, 2.0614128, 4.2246304, 6.0238333,
# 10.107266, 8.601249, 8.808282], dtype=np.float32)
# datas['mean'] = [0.5, 0.5, 0.5]
# datas['std'] = [0.2, 0.2, 0.2]
# define loss function, respectively
weight = torch.from_numpy(datas['classWeights'])
if args.dataset == 'pollen':
weight = torch.tensor([1., 1.])
if args.dataset == 'camvid':
criteria = CrossEntropyLoss2d(weight=weight,
ignore_label=args.ignore_label)
elif args.dataset == 'camvid' and args.use_label_smoothing:
criteria = CrossEntropyLoss2dLabelSmooth(
weight=weight, ignore_label=args.ignore_label)
elif args.dataset == 'cityscapes' and args.use_ohem:
min_kept = int(args.batch_size // len(args.gpus) * h * w // 16)
criteria = ProbOhemCrossEntropy2d(use_weight=True,
ignore_label=args.ignore_label,
thresh=0.7,
min_kept=min_kept)
elif args.dataset == 'cityscapes' and args.use_label_smoothing:
criteria = CrossEntropyLoss2dLabelSmooth(
weight=weight, ignore_label=args.ignore_label)
elif args.dataset == 'cityscape' and args.use_lovaszsoftmax:
criteria = LovaszSoftmax(ignore_index=args.ignore_label)
elif args.dataset == 'cityscapes' and args.use_focal:
criteria = FocalLoss2d(weight=weight, ignore_index=args.ignore_label)
elif args.dataset == 'seed':
criteria = CrossEntropyLoss2d(weight=weight,
ignore_label=args.ignore_label)
elif args.dataset == 'remote':
criteria = CrossEntropyLoss2d(weight=weight,
ignore_label=args.ignore_label)
elif args.dataset == 'remote' and args.use_ohem:
min_kept = int(args.batch_size // len(args.gpus) * h * w // 16)
criteria = ProbOhemCrossEntropy2d(use_weight=True,
ignore_label=args.ignore_label,
thresh=0.7,
min_kept=min_kept)
elif args.dataset == 'remote' and args.use_label_smoothing:
criteria = CrossEntropyLoss2dLabelSmooth(
weight=weight, ignore_label=args.ignore_label)
elif args.dataset == 'remote' and args.use_lovaszsoftmax:
criteria = LovaszSoftmax(ignore_index=args.ignore_label)
elif args.dataset == 'remote' and args.use_focal:
criteria = FocalLoss2d(weight=weight, ignore_index=args.ignore_label)
else:
criteria = CrossEntropyLoss2d(weight=weight,
ignore_label=args.ignore_label)
if args.cuda:
criteria = criteria.cuda()
if torch.cuda.device_count() > 1:
print("torch.cuda.device_count()=", torch.cuda.device_count())
args.gpu_nums = torch.cuda.device_count()
model = nn.DataParallel(model).cuda() # multi-card data parallel
else:
args.gpu_nums = 1
print("single GPU for training")
model = model.cuda() # 1-card data parallel
args.savedir = (args.savedir + args.dataset + '/' + args.model + 'bs' +
str(args.batch_size) + 'gpu' + str(args.gpu_nums) + "_" +
str(args.train_type) + '/')
if not os.path.exists(args.savedir):
os.makedirs(args.savedir)
start_epoch = 0
# continue training
if args.resume:
if os.path.isfile(args.resume):
checkpoint = torch.load(args.resume)
start_epoch = checkpoint['epoch']
model.load_state_dict(checkpoint['model'])
# model.load_state_dict(convert_state_dict(checkpoint['model']))
print("=====> loaded checkpoint '{}' (epoch {})".format(
args.resume, checkpoint['epoch']))
else:
print("=====> no checkpoint found at '{}'".format(args.resume))
model.train()
cudnn.benchmark = True
# cudnn.deterministic = True ## my add
logFileLoc = args.savedir + args.logFile
if os.path.isfile(logFileLoc):
logger = open(logFileLoc, 'a')
else:
logger = open(logFileLoc, 'w')
logger.write("Parameters: %s Seed: %s" %
(str(total_parameters), GLOBAL_SEED))
logger.write("\n%s\t\t%s\t%s\t%s" %
('Epoch', 'Loss(Tr)', 'mIOU (val)', 'lr'))
logger.flush()
# define optimization strategy
if args.optim == 'sgd':
optimizer = torch.optim.SGD(filter(lambda p: p.requires_grad,
model.parameters()),
lr=args.lr,
momentum=0.9,
weight_decay=1e-4)
elif args.optim == 'adam':
optimizer = torch.optim.Adam(filter(lambda p: p.requires_grad,
model.parameters()),
lr=args.lr,
betas=(0.9, 0.999),
eps=1e-08,
weight_decay=1e-4)
elif args.optim == 'radam':
optimizer = RAdam(filter(lambda p: p.requires_grad,
model.parameters()),
lr=args.lr,
betas=(0.90, 0.999),
eps=1e-08,
weight_decay=1e-4)
elif args.optim == 'ranger':
optimizer = Ranger(filter(lambda p: p.requires_grad,
model.parameters()),
lr=args.lr,
betas=(0.95, 0.999),
eps=1e-08,
weight_decay=1e-4)
elif args.optim == 'adamw':
optimizer = AdamW(filter(lambda p: p.requires_grad,
model.parameters()),
lr=args.lr,
betas=(0.9, 0.999),
eps=1e-08,
weight_decay=1e-4)
lossTr_list = []
epoches = []
mIOU_val_list = []
print('=====> beginning training')
for epoch in range(start_epoch, args.max_epochs):
# training
lossTr, lr = train(args, trainLoader, model, criteria, optimizer,
epoch)
lossTr_list.append(lossTr)
# validation
if epoch % 2 == 0 or epoch == (args.max_epochs - 1):
epoches.append(epoch)
mIOU_val, per_class_iu = val(args, valLoader, model)
mIOU_val_list.append(mIOU_val)
# record train information
logger.write("\n%d\t\t%.4f\t\t%.4f\t\t%.7f" %
(epoch, lossTr, mIOU_val, lr))
logger.flush()
print("Epoch : " + str(epoch) + ' Details')
print(
"Epoch No.: %d\tTrain Loss = %.4f\t mIOU(val) = %.4f\t lr= %.6f\n"
% (epoch, lossTr, mIOU_val, lr))
else:
# record train information
logger.write("\n%d\t\t%.4f\t\t\t\t%.7f" % (epoch, lossTr, lr))
logger.flush()
print("Epoch : " + str(epoch) + ' Details')
print("Epoch No.: %d\tTrain Loss = %.4f\t lr= %.6f\n" %
(epoch, lossTr, lr))
# save the model
model_file_name = args.savedir + '/model_' + str(epoch + 1) + '.pth'
state = {"epoch": epoch + 1, "model": model.state_dict()}
# Individual Setting for save model !!!
if args.dataset == 'camvid':
torch.save(state, model_file_name)
elif args.dataset == 'cityscapes':
if epoch >= args.max_epochs - 10:
torch.save(state, model_file_name)
elif not epoch % 50:
torch.save(state, model_file_name)
elif args.dataset == 'seed':
torch.save(state, model_file_name)
else:
torch.save(state, model_file_name)
# draw plots for visualization
if epoch % 5 == 0 or epoch == (args.max_epochs - 1):
# Plot the figures per 50 epochs
fig1, ax1 = plt.subplots(figsize=(11, 8))
ax1.plot(range(start_epoch, epoch + 1), lossTr_list)
ax1.set_title("Average training loss vs epochs")
ax1.set_xlabel("Epochs")
ax1.set_ylabel("Current loss")
plt.savefig(args.savedir + "loss_vs_epochs.png")
plt.clf()
fig2, ax2 = plt.subplots(figsize=(11, 8))
ax2.plot(epoches, mIOU_val_list, label="Val IoU")
ax2.set_title("Average IoU vs epochs")
ax2.set_xlabel("Epochs")
ax2.set_ylabel("Current IoU")
plt.legend(loc='lower right')
plt.savefig(args.savedir + "iou_vs_epochs.png")
plt.close('all')
logger.close()
def train_model(args):
"""
args:
args: global arguments
"""
h, w = map(int, args.input_size.split(','))
input_size = (h, w)
print("input size:{}".format(input_size))
print(args)
if args.cuda:
print("use gpu id: '{}'".format(args.gpus))
os.environ["CUDA_VISIBLE_DEVICES"] = args.gpus
if not torch.cuda.is_available():
raise Exception(
"No GPU found or Wrong gpu id, please run without --cuda")
# set the seed
setup_seed(GLOBAL_SEED)
print("set Global Seed: ", GLOBAL_SEED)
cudnn.enabled = True
print("building network")
# build the model and initialization
model = build_model(args.model, num_classes=args.classes)
init_weight(model,
nn.init.kaiming_normal_,
nn.BatchNorm2d,
1e-3,
0.1,
mode='fan_in')
print("computing network parameters and FLOPs")
total_paramters = netParams(model)
print("the number of parameters: %d ==> %.2f M" %
(total_paramters, (total_paramters / 1e6)))
# load data and data augmentation
datas, trainLoader, valLoader = build_dataset_train(
args.dataset, input_size, args.batch_size, args.train_type,
args.random_scale, args.random_mirror, args.num_workers)
args.per_iter = len(trainLoader)
args.max_iter = args.max_epochs * args.per_iter
print('Dataset statistics')
print("data['classWeights']: ", datas['classWeights'])
print('mean and std: ', datas['mean'], datas['std'])
# define loss function, respectively
weight = torch.from_numpy(datas['classWeights'])
if args.dataset == 'camvid':
criteria = CrossEntropyLoss2d(weight=weight, ignore_label=ignore_label)
elif args.dataset == 'camvid' and args.use_label_smoothing:
criteria = CrossEntropyLoss2dLabelSmooth(weight=weight,
ignore_label=ignore_label)
elif args.dataset == 'cityscapes' and args.use_ohem:
min_kept = int(args.batch_size // len(args.gpus) * h * w // 16)
criteria = ProbOhemCrossEntropy2d(use_weight=True,
ignore_label=ignore_label,
thresh=0.7,
min_kept=min_kept)
elif args.dataset == 'cityscapes' and args.use_label_smoothing:
criteria = CrossEntropyLoss2dLabelSmooth(weight=weight,
ignore_label=ignore_label)
elif args.dataset == 'cityscapes' and args.use_lovaszsoftmax:
criteria = LovaszSoftmax(ignore_index=ignore_label)
elif args.dataset == 'cityscapes' and args.use_focal:
criteria = FocalLoss2d(weight=weight, ignore_index=ignore_label)
elif args.dataset == 'paris':
criteria = CrossEntropyLoss2d(weight=weight, ignore_label=ignore_label)
else:
raise NotImplementedError(
"This repository now supports two datasets: cityscapes and camvid, %s is not included"
% args.dataset)
if args.cuda:
criteria = criteria.cuda()
if torch.cuda.device_count() > 1:
print("torch.cuda.device_count()=", torch.cuda.device_count())
args.gpu_nums = torch.cuda.device_count()
model = nn.DataParallel(model).cuda() # multi-card data parallel
else:
args.gpu_nums = 1
print("single GPU for training")
model = model.cuda() # 1-card data parallel
args.savedir = (args.savedir + args.dataset + '/' + args.model + 'bs' +
str(args.batch_size) + 'gpu' + str(args.gpu_nums) + "_" +
str(args.train_type) + '/')
if not os.path.exists(args.savedir):
os.makedirs(args.savedir)
with open(args.savedir + 'args.txt', 'w') as f:
f.write('mean:{}\nstd:{}\n'.format(datas['mean'], datas['std']))
f.write("Parameters: {} Seed: {}\n".format(str(total_paramters),
GLOBAL_SEED))
f.write(str(args))
start_epoch = 0
# continue training
if args.resume:
if os.path.isfile(args.resume):
checkpoint = torch.load(args.resume)
start_epoch = checkpoint['epoch']
model.load_state_dict(checkpoint['model'])
# model.load_state_dict(convert_state_dict(checkpoint['model']))
print("loaded checkpoint '{}' (epoch {})".format(
args.resume, checkpoint['epoch']))
else:
print("no checkpoint found at '{}'".format(args.resume))
model.train()
cudnn.benchmark = True
# cudnn.deterministic = True ## my add
# initialize the early_stopping object
early_stopping = EarlyStopping(patience=50)
logFileLoc = args.savedir + args.logFile
if os.path.isfile(logFileLoc):
logger = open(logFileLoc, 'a')
else:
logger = open(logFileLoc, 'w')
logger.write("%s\t%s\t\t%s\t%s\t%s" %
('Epoch', ' lr', 'Loss(Tr)', 'Loss(Val)', 'mIOU(Val)'))
logger.flush()
# define optimization strategy
if args.optim == 'sgd':
optimizer = torch.optim.SGD(filter(lambda p: p.requires_grad,
model.parameters()),
lr=args.lr,
momentum=0.9,
weight_decay=1e-4)
elif args.optim == 'adam':
optimizer = torch.optim.Adam(filter(lambda p: p.requires_grad,
model.parameters()),
lr=args.lr,
betas=(0.9, 0.999),
eps=1e-08,
weight_decay=1e-4)
elif args.optim == 'radam':
optimizer = RAdam(filter(lambda p: p.requires_grad,
model.parameters()),
lr=args.lr,
betas=(0.90, 0.999),
eps=1e-08,
weight_decay=1e-4)
elif args.optim == 'ranger':
optimizer = Ranger(filter(lambda p: p.requires_grad,
model.parameters()),
lr=args.lr,
betas=(0.95, 0.999),
eps=1e-08,
weight_decay=1e-4)
elif args.optim == 'adamw':
optimizer = AdamW(filter(lambda p: p.requires_grad,
model.parameters()),
lr=args.lr,
betas=(0.9, 0.999),
eps=1e-08,
weight_decay=1e-4)
lossTr_list = []
epoches = []
mIOU_val_list = []
lossVal_list = []
print('>>>>>>>>>>>beginning training>>>>>>>>>>>')
for epoch in range(start_epoch, args.max_epochs):
# training
lossTr, lr = train(args, trainLoader, model, criteria, optimizer,
epoch)
lossTr_list.append(lossTr)
# validation
if epoch % args.val_miou_epochs == 0:
epoches.append(epoch)
val_loss, mIOU_val, per_class_iu = val(args, valLoader, criteria,
model, epoch)
mIOU_val_list.append(mIOU_val)
lossVal_list.append(val_loss.item())
# record train information
logger.write(
"\n%d\t%.6f\t%.4f\t\t%.4f\t%0.4f\t %s" %
(epoch, lr, lossTr, val_loss, mIOU_val, str(per_class_iu)))
logger.flush()
print(
"Epoch %d\tlr= %.6f\tTrain Loss = %.4f\tVal Loss = %.4f\tmIOU(val) = %.4f\tper_class_iu= %s\n"
% (epoch, lr, lossTr, val_loss, mIOU_val, str(per_class_iu)))
else:
# record train information
val_loss = val(args, valLoader, criteria, model, epoch)
lossVal_list.append(val_loss.item())
logger.write("\n%d\t%.6f\t%.4f\t\t%.4f" %
(epoch, lr, lossTr, val_loss))
logger.flush()
print("Epoch %d\tlr= %.6f\tTrain Loss = %.4f\tVal Loss = %.4f\n" %
(epoch, lr, lossTr, val_loss))
# save the model
model_file_name = args.savedir + '/model_' + str(epoch) + '.pth'
state = {"epoch": epoch, "model": model.state_dict()}
# Individual Setting for save model
if epoch >= args.max_epochs - 10:
torch.save(state, model_file_name)
elif epoch % 10 == 0:
torch.save(state, model_file_name)
# draw plots for visualization
if os.path.isfile(args.savedir + "loss.png"):
f = open(args.savedir + 'log.txt', 'r')
next(f)
epoch_list = []
lossTr_list = []
lossVal_list = []
for line in f.readlines():
epoch_list.append(line.strip().split()[0])
lossTr_list.append(line.strip().split()[2])
lossVal_list.append(line.strip().split()[3])
assert len(epoch_list) == len(lossTr_list) == len(lossVal_list)
fig1, ax1 = plt.subplots(figsize=(11, 8))
ax1.plot(range(0, epoch + 1), lossTr_list, label='Train_loss')
ax1.plot(range(0, epoch + 1), lossVal_list, label='Val_loss')
ax1.set_title("Average training loss vs epochs")
ax1.set_xlabel("Epochs")
ax1.set_ylabel("Current loss")
ax1.legend()
plt.savefig(args.savedir + "loss.png")
plt.clf()
else:
fig1, ax1 = plt.subplots(figsize=(11, 8))
ax1.plot(range(0, epoch + 1), lossTr_list, label='Train_loss')
ax1.plot(range(0, epoch + 1), lossVal_list, label='Val_loss')
ax1.set_title("Average training loss vs epochs")
ax1.set_xlabel("Epochs")
ax1.set_ylabel("Current loss")
ax1.legend()
plt.savefig(args.savedir + "loss.png")
plt.clf()
fig2, ax2 = plt.subplots(figsize=(11, 8))
ax2.plot(epoches, mIOU_val_list, label="Val IoU")
ax2.set_title("Average IoU vs epochs")
ax2.set_xlabel("Epochs")
ax2.set_ylabel("Current IoU")
ax2.legend()
plt.savefig(args.savedir + "mIou.png")
plt.close('all')
early_stopping.monitor(monitor=val_loss)
if early_stopping.early_stop:
print("Early stopping and Save checkpoint")
if not os.path.exists(model_file_name):
torch.save(state, model_file_name)
break
logger.close()
