def main(train_args):
trainset = "/mnt/iusers01/eee01/mchiwml4/CamVid/train"
validset = "/mnt/iusers01/eee01/mchiwml4/CamVid/val"
transform = transforms.Compose(
[transforms.ToTensor(),
transforms.Normalize(mean, std)])
train_joint_transform = jointtransform.Compose(
[jointtransform.RandomHorizontallyFlip()])
train_dataset = DataLoader(Loaddata(trainset,
transform=transform,
target_transform=MaskToTensor(),
joint_transform=train_joint_transform),
batch_size=train_args.batch_size,
shuffle=True,
num_workers=8)
valid_dataset = DataLoader(Loaddata(validset,
transform=transform,
target_transform=MaskToTensor()),
batch_size=1,
shuffle=True,
num_workers=8)
label_num = 11
model = ducmodel.ResNetDUCHDC(label_num)
model = model.cuda()
weight = torch.Tensor(class_weight).cuda()
criterion = CrossEntropyLoss2d(weight=weight).cuda()
lr_ = train_args.lr
optimizer = optim.SGD([{
'params': get_1x_lr_params(model),
'lr': lr_
}, {
'params': get_10x_lr_params(model),
'lr': lr_ * 10
}],
momentum=train_args.momentum,
weight_decay=train_args.weight_decay)
# optimizer.param_groups[0]['lr'] = lr_
# optimizer.param_groups[1]['lr'] = lr_*10
if train_args.load_param is not None:
model.load_state_dict(torch.load(train_args.load_param))
if train_args.load_optim is not None:
optimizer.load_state_dict(torch.load(train_args.load_optim))
max_epochs = train_args.epoch_num
savefolder = train_args.save_folder
train_model(model, criterion, optimizer, lr_, max_epochs, train_dataset,
valid_dataset, savefolder)
def main(train_args):
trainset = "/mnt/iusers01/eee01/mchiwml4/CamVid/train"
validset = "/mnt/iusers01/eee01/mchiwml4/CamVid/val"
transform = transforms.Compose(
[transforms.ToTensor(),
transforms.Normalize(mean, std)])
train_joint_transform = jointtransform.Compose([
jointtransform.RandomCrop(224),
jointtransform.RandomHorizontallyFlip()
])
train_dataset = DataLoader(Loaddata(trainset,
transform=transform,
target_transform=MaskToTensor(),
joint_transform=train_joint_transform),
batch_size=train_args.batch_size,
shuffle=True,
num_workers=8)
valid_dataset = DataLoader(Loaddata(validset,
transform=transform,
target_transform=MaskToTensor()),
batch_size=1,
shuffle=True,
num_workers=8)
label_num = 11
model = fcdensenetmodel.FCDenseNet103(label_num)
model = model.cuda()
weight = torch.Tensor(class_weight).cuda()
criterion = CrossEntropyLoss2d(weight=weight).cuda()
lr_ = train_args.lr
optimizer = optim.RMSprop(model.parameters(),
lr=lr_,
weight_decay=train_args.weight_decay)
exp_lr_scheduler = optim.lr_scheduler.ExponentialLR(optimizer, gamma=0.995)
if train_args.load_param is not None:
model.load_state_dict(torch.load(train_args.load_param))
if train_args.load_optim is not None:
optimizer.load_state_dict(torch.load(train_args.load_optim))
max_epochs = train_args.epoch_num
savefolder = train_args.save_folder
train_model(model, criterion, optimizer, exp_lr_scheduler, max_epochs,
train_dataset, valid_dataset, savefolder)
def main(test_args):
testset = "/mnt/iusers01/eee01/mchiwml4/CamVid/test"
transform = transforms.Compose(
[transforms.ToTensor(),
transforms.Normalize(mean, std)])
test_dataset = DataLoader(Loaddata(testset,
transform=transform,
target_transform=MaskToTensor()),
batch_size=1,
shuffle=False,
num_workers=8)
label_num = 11
model = linknetmodel.linknet(label_num)
model = model.cuda()
model.load_state_dict(torch.load(test_args.load_param))
model.eval()
total = np.zeros((label_num, ))
running_metrics = runningScore(label_num)
time_elapsed = 0
for j, data in enumerate(test_dataset):
since = time.time()
inputs, labels = data
inputs = Variable(inputs.cuda(), volatile=True)
outputs = model(inputs)
time_elapsed += time.time() - since
pred = outputs.data.max(1)[1].cpu().numpy()
gt = labels.numpy()
running_metrics.update(gt, pred)
for i in range(label_num):
mask = gt == i # ground truth mask of class i
total[i] += np.sum(
mask) # total number of pixels of class i (tp+fn)
print('Inference speed: {:.0f}ms, {:.0f}fps '.format(
time_elapsed / len(test_dataset) * 1000,
1 / (time_elapsed / len(test_dataset))))
score, class_iou, class_acc = running_metrics.get_scores()
for k, v in score.items():
print(k, v)
print('class iou: ')
for i in range(label_num):
print(i, class_iou[i])
print('class acc: ')
for i in range(label_num):
print(i, class_acc[i])
print('number of pixels:')
print(total)
def main(test_args):
testset="/mnt/iusers01/eee01/mchiwml4/CamVid/test"
transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize(mean, std)
])
test_dataset = DataLoader(
Loaddata(testset, transform=transform, target_transform=MaskToTensor()),
batch_size=1, shuffle=False, num_workers=8)
label_num=11
model = ducmodel.ResNetDUCHDC(label_num)
model=model.cuda()
model.load_state_dict(torch.load(test_args.load_param))
model.eval()
total=np.zeros((label_num,))
running_metrics = runningScore(label_num)
for j, data in enumerate(test_dataset):
inputs, labels = data
inputs = Variable(inputs.cuda())
outputs = model(inputs)
pred = outputs.data.max(1)[1].cpu().numpy()
gt = labels.numpy()
running_metrics.update(gt, pred)
print(j)
for i in range(label_num):
mask=gt==i # ground truth mask of class i
total[i]+=np.sum(mask)
score, class_iou, class_acc = running_metrics.get_scores()
for k, v in score.items():
print(k, v)
print('class iou: ')
for i in range(label_num):
print(i, class_iou[i])
print('class acc: ')
for i in range(label_num):
print(i, class_acc[i])
print('number of pixels:')
print(total)
def main(test_args):
testset = "/mnt/iusers01/eee01/mchiwml4/CamVid/test"
transform = transforms.Compose(
[transforms.ToTensor(),
transforms.Normalize(mean, std)])
test_dataset = DataLoader(Loaddata(testset,
transform=transform,
target_transform=MaskToTensor()),
batch_size=1,
shuffle=False,
num_workers=8)
label_num = 11
model = deeplab_v2.Deeplab_Resnet(label_num)
model = model.cuda()
model.load_state_dict(torch.load(test_args.load_param))
model.eval()
total = np.zeros((label_num, ))
running_metrics = runningScore(label_num)
for j, data in enumerate(test_dataset):
inputs, labels = data
inputs = Variable(inputs.cuda())
outputs = model(inputs)
H = inputs.size()[2]
W = inputs.size()[3]
interp_resize = nn.Upsample(size=(int(H), int(W)), mode='bilinear')
output = interp_resize(outputs[3])
output = F.softmax(output, dim=1)
output = output.data.cpu().numpy()
if test_args.crf:
crf_output = np.zeros(output.shape)
images = inputs.data.cpu().numpy().astype(np.uint8)
for i, (image, prob_map) in enumerate(zip(images, output)):
image = image.transpose(1, 2, 0)
crf_output[i] = dense_crf(image, prob_map)
output = crf_output
pred = np.argmax(output, axis=1)
gt = labels.numpy()
running_metrics.update(gt, pred)
for i in range(label_num):
mask = gt == i # ground truth mask of class i
total[i] += np.sum(
mask) # total number of pixels of class i (tp+fn)
score, class_iou, class_acc = running_metrics.get_scores()
for k, v in score.items():
print(k, v)
print('class iou: ')
for i in range(label_num):
print(i, class_iou[i])
print('class acc: ')
for i in range(label_num):
print(i, class_acc[i])
print('number of pixels:')
print(total)
trainset="/mnt/iusers01/eee01/mchiwml4/CamVid/train"
validset="/mnt/iusers01/eee01/mchiwml4/CamVid/val"
testset="/mnt/iusers01/eee01/mchiwml4/CamVid/test"
transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize(mean, std)
])
train_joint_transform = jointtransform.Compose([
jointtransform.RandomHorizontallyFlip()
])
train_dataset = DataLoader(
Loaddata(trainset, transform=transform, target_transform=MaskToTensor(), joint_transform=train_joint_transform),
batch_size=4, shuffle=True, num_workers=8)
valid_dataset = DataLoader(
Loaddata(validset, transform=transform, target_transform=MaskToTensor()),
batch_size=1, shuffle=True, num_workers=8)
train_size = len(Loaddata(trainset))
valid_size = len(Loaddata(validset))
def get_1x_lr_params(model):
b = []
b.append(model.down1.parameters())
b.append(model.down2.parameters())
b.append(model.down3.parameters())
import torch.nn.functional as F
import numpy as np
import utils.jointtransform as jointtransform
trainset = "/mnt/iusers01/eee01/mchiwml4/CamVid/train"
validset = "/mnt/iusers01/eee01/mchiwml4/CamVid/val"
testset = "/mnt/iusers01/eee01/mchiwml4/CamVid/test"
transform = transforms.Compose(
[transforms.ToTensor(),
transforms.Normalize(mean, std)])
train_joint_transform = jointtransform.Compose(
[jointtransform.RandomHorizontallyFlip()])
train_dataset = DataLoader(Loaddata(trainset,
transform=transform,
target_transform=MaskToTensor(),
joint_transform=train_joint_transform),
batch_size=4,
shuffle=True,
num_workers=8)
valid_dataset = DataLoader(Loaddata(validset,
transform=transform,
target_transform=MaskToTensor()),
batch_size=1,
shuffle=True,
num_workers=8)
train_size = len(Loaddata(trainset))
valid_size = len(Loaddata(validset))
test_size = len(Loaddata(testset))