def CalculateRay(self, screen, x, y):
# https://learnopengl.com/Lighting/Basic-Lighting for shading
steps = 0
depth = 0
max_steps = 1
x_angle = MapVals(x, 0, self.WIDTH, -self.FOV / 2,
self.FOV / 2) * self.ASP # the x angle of this ray
y_angle = MapVals(y, 0, self.HEIGHT, -self.FOV / 2,
self.FOV / 2) # the y angle of this ray
direction = np.array([math.sin(x_angle),
math.sin(y_angle),
1]) # turning these angles into a vector
for i in range(0, self.MARCH_STEPS):
point = (self.camera.transform.xyz + depth) * direction
dist = self.SceneSDF(point)
if dist < self.EPSILON:
# INSIDE SURFACE
norm = self.SurfaceNormalEstimate(point)
ambiend_rgb = [0.1, 0.1, 0.1]
diff_rgb = [0, 0, 0]
spec_rgb = [0, 0, 0]
sum_rgb = [0, 0, 0]
light_level = len(self.lights)
if (len(self.lights) == 0):
light_level = 1
for light in self.lights:
light_dir = Normalize(light.transform.xyz - point)
# Calculate diffuse intensity
diff_intensity = np.dot(light_dir, norm)
# Calculate specular intensity
spec_power = 32
spec_strength = 1
reflect_dir = Reflect(light_dir, norm)
spec = math.pow(np.dot(light_dir, reflect_dir), spec_power)
spec_intensity = spec_strength * spec
for i in range(0, 3):
diff_rgb[i] += (diff_intensity * light.colour[i])
spec_rgb[i] += (spec_intensity * light.colour[i])
for i in range(0, 3):
sum = ambiend_rgb[i] + diff_rgb[i] + spec_rgb[i]
if sum < 0:
sum = 0
#sum_rgb[i]=int(MapVals(sum, 0, light_level, 0, 255))
sum_rgb[i] = int(MapVals(sum, 0, 1, 0, 255))
if sum_rgb[i] > 255:
sum_rgb[i] = 255
return (sum_rgb[0], sum_rgb[1], sum_rgb[2])
depth += dist
steps += 1
if depth >= self.END:
if steps > max_steps:
max_steps = steps
# Gone too far
# Apply edge glow
if steps >= self.GLOW_THRESHOLD:
col = MapVals(steps, self.GLOW_THRESHOLD, max_steps + 1,
self.BACKGROUND_COLOUR[0], 255)
return (col, col, col)
return self.BACKGROUND_COLOUR
def normalizeVariable(self):
# Get selected variable in preview Tab widget
selected_var_idx = [
self.tableWidget_signal_preview.selectedIndexes()[i].column()
for i in range(
len(self.tableWidget_signal_preview.selectedIndexes()))
]
selected_var_idx = sorted(set(selected_var_idx))
selected_variables = []
for i in selected_var_idx:
if i != 0: #ignore if TimeIndex is selected
selected_variable = self.my_signal._signal_data.columns.values[
i - 1]
min_value, min_ok = QtGui.QInputDialog.getInt(
self,
'Normalize Variable',
"Give a minimal value for the normalization of " +
selected_variable + " :",
value=0)
if min_ok:
max_value, max_ok = QtGui.QInputDialog.getInt(
self,
'Normalize Variable',
"Give a maximal value for the normalization of " +
selected_variable + " :",
value=min_value + 1)
if max_ok:
self.my_signal._signal_data[
selected_variable] = Normalize.Normalize(
pd.DataFrame(self.my_signal.
_signal_data[selected_variable]),
min_value, max_value)
self.displaySignalPreview()
def todfa():
states_json = eval(request.values['states'])
states = {}
for s in states_json:
isInitial, isFinal = False, False
if s.get("isInitial"):
isInitial = True
if s.get("isFinal"):
isFinal = True
states.update(
{s['id']: state(s['label'], isFinal=isFinal, isInitial=isInitial)})
transitions_json = eval(request.values['transitions'])
for t in transitions_json:
states[t['from']].connect(states[t['to']], t['label'], t['pop'],
t['push'])
machine = pda()
for s in states.values():
machine.add_state(s)
#return (str(machine).replace("\n","<br>"))
normal_machine = Normalize(machine)
grammer = MakeCFG(normal_machine)
states_json = []
transitions_json = []
transition_id = 0
print(machine)
for s in normal_machine.states:
new_state = {"id": s.name, "label": s.name}
if s.isFinal:
new_state.update({"isFinal": 1})
if s.isInitial:
new_state.update({"isInitial": 1})
states_json.append(new_state)
tmp = s.get_neighbours()
for t in tmp:
for next_s in tmp[t]:
new_edge = {
"id": str(transition_id),
"from": s.name,
"to": next_s,
"arrows": "to"
}
tmp0 = []
for i in [0, 1, 2]:
if t[i] == "":
tmp0.append(chr(955))
else:
tmp0.append(str(t[i]))
new_edge.update(
{"label": tmp0[0] + ", " + tmp0[1] + " --> " + tmp0[2]})
transitions_json.append(new_edge)
transition_id += 1
good_grammer = ""
for i in grammer:
good_grammer += "<br>" + str(i) + " ---> " + "|".join(
["".join(j) for j in grammer[i]])
return render_template("show_dfa.html",
states_json=states_json,
transitions_json=transitions_json,
grammer=good_grammer)
def create_vgg_model(style_layers=[0, 7, 14, 27, 40],
content_layers=[30],
style_weight=[1.0, 1.0, 1.0, 1.0, 1.0],
content_weight=[1.0]):
vgg = freeze_until(
models.vgg19_bn(pretrained=True).to(self.device).features, 99999)
style_losses = []
content_losses = []
calc_loss = False
model = nn.Sequential(
Normalize(
torch.tensor([0.485, 0.456, 0.406]).to(self.device),
torch.tensor([0.229, 0.224, 0.225]).to(self.device)))
for i, layer in enumerate(vgg.children()):
if (isinstance(layer, nn.ReLU)):
model.add_module(str(i), nn.ReLU(inplace=False))
else:
model.add_module(str(i), layer)
if (i in style_layers):
style_loss_layer = StyleLoss(model(s_img).detach())
model.add_module('Style-' + str(i), style_loss_layer)
style_layers.remove(i)
style_losses.append(style_loss_layer)
if (i in content_layers):
content_loss_layer = ContentLoss(model(c_img).detach())
model.add_module('Content-' + str(i), content_loss_layer)
content_layers.remove(i)
content_losses.append(content_loss_layer)
if (len(style_layers) == 0 and len(content_layers) == 0):
break
return model
def main():
args = get_args()
root_dir = args.root_dir
imgs = list(os.walk(root_dir))[0][2]
save_dir = args.save_dir
num_classes = 100 # CIFAR100
model = ResNet.resnet(arch='resnet50', pretrained=False, num_classes=num_classes,
use_att=args.use_att, att_mode=args.att_mode)
#model = nn.DataParallel(model)
#print(model)
if args.resume:
if os.path.isfile(args.resume):
print(f'=> loading checkpoint {args.resume}')
checkpoint = torch.load(args.resume)
best_acc5 = checkpoint['best_acc5']
model.load_state_dict(checkpoint['state_dict'], strict=False)
print(f"=> loaded checkpoint {args.resume} (epoch {checkpoint['epoch']})")
print(f'=> best accuracy {best_acc5}')
else:
print(f'=> no checkpoint found at {args.resume}')
model_dict = get_model_dict(model, args.type)
normalizer = Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
for img_name in imgs:
img_path = os.path.join(root_dir, img_name)
pil_img = PIL.Image.open(img_path)
torch_img = torch.from_numpy(np.asarray(pil_img))
torch_img = torch_img.permute(2, 0, 1).unsqueeze(0)
torch_img = torch_img.float().div(255)
torch_img = F.interpolate(torch_img, size=(224, 224), mode='bilinear', align_corners=False)
normalized_torch_img = normalizer(torch_img)
gradcam = GradCAM(model_dict, True)
gradcam_pp = GradCAMpp(model_dict, True)
mask, _ = gradcam(normalized_torch_img)
heatmap, result = visualize_cam(mask, torch_img)
mask_pp, _ = gradcam_pp(normalized_torch_img)
heatmap_pp, result_pp = visualize_cam(mask_pp, torch_img)
images = torch.stack([torch_img.squeeze().cpu(), heatmap, heatmap_pp, result, result_pp], 0)
images = make_grid(images, nrow=1)
if args.use_att:
save_dir = os.path.join(args.save_dir, 'att')
else:
save_dir = os.path.join(args.save_dir, 'no_att')
os.makedirs(save_dir, exist_ok=True)
output_name = img_name
output_path = os.path.join(save_dir, output_name)
save_image(images, output_path)
def SurfaceNormalEstimate(self, pos):
return Normalize(
np.array([
self.SceneSDF(np.array(
[pos[0] + self.EPSILON, pos[1], pos[2]])) - self.SceneSDF(
np.array([pos[0] - self.EPSILON, pos[1], pos[2]])),
self.SceneSDF(np.array(
[pos[0], pos[1] + self.EPSILON, pos[2]])) - self.SceneSDF(
np.array([pos[0], pos[1] - self.EPSILON, pos[2]])),
self.SceneSDF(np.array(
[pos[0], pos[1], pos[2] + self.EPSILON])) - self.SceneSDF(
np.array([pos[0], pos[1], pos[2] - self.EPSILON]))
]))
def generate_saliency_map(img, img_name):
start = time.time()
normalizer = Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
torch_img = torch.from_numpy(np.asarray(img)).permute(
2, 0, 1).unsqueeze(0).float().div(255)
torch_img = F.upsample(torch_img,
size=(512, 512),
mode='bilinear',
align_corners=False)
normed_torch_img = normalizer(torch_img)
resnet = models.resnet101(pretrained=True)
resnet.eval()
cam_dict = dict()
model_dict = dict(type='resnet',
arch=resnet,
layer_name='layer4',
input_size=(512, 512))
gradcam = GradCAM(model_dict, True)
gradcam_pp = GradCAMpp(model_dict, True)
images = []
mask, _ = gradcam(normed_torch_img)
heatmap, result = visualize_cam(mask, torch_img)
mask_pp, _ = gradcam_pp(normed_torch_img)
heatmap_pp, result_pp = visualize_cam(mask_pp, torch_img)
images.append(
torch.stack([
torch_img.squeeze().cpu(), heatmap, heatmap_pp, result, result_pp
], 0))
images = make_grid(torch.cat(images, 0), nrow=1)
# Only going to use result_pp
output_dir = 'outputs'
os.makedirs(output_dir, exist_ok=True)
output_name = img_name
output_path = os.path.join(output_dir, output_name)
save_image(result_pp, output_path)
end = time.time()
duration = round(end - start, 2)
return output_path
invTransform = Denormalize(opt.normalizeMean, opt.normalizeNorm)
# set model
model = createModel(opt)
model.setup(opt)
# set dataloader
if opt.augment:
transformList = [
Resize(opt.loadSize),
RandomCrop(opt.fineSize),
RandomRotation(opt.rotate),
ToTensor(),
Normalize(opt.normalizeMean, opt.normalizeNorm),
RandomHorizontalFlip(),
]
else:
transformList = [
Resize(opt.loadSize),
ToTensor(),
Normalize(opt.normalizeMean, opt.normalizeNorm)
]
transform = Compose(transformList)
datasetA = createDataset([opt.datasetA], transform=transform, outputFile=False)
datasetB = createDataset([opt.datasetB], transform=transform, outputFile=False)
def setup(self):
pkl_dir = self.config.split_dir
with open(os.path.join(pkl_dir, "splits.pkl"), 'rb') as f:
splits = pickle.load(f)
tr_keys = splits[self.config.fold]['train']
val_keys = splits[self.config.fold]['val']
test_keys = splits[self.config.fold]['test']
print("pkl_dir: ", pkl_dir)
print("tr_keys: ", tr_keys)
print("val_keys: ", val_keys)
print("test_keys: ", test_keys)
self.device = torch.device(
self.config.device if torch.cuda.is_available() else "cpu")
task = self.config.dataset_name
self.train_data_loader = torch.utils.data.DataLoader(
NucleusDataset(self.config.data_root_dir,
train=True,
transform=transforms.Compose([
Normalize(),
Rescale(self.config.patch_size),
ToTensor()
]),
target_transform=transforms.Compose([
Normalize(),
Rescale(self.config.patch_size),
ToTensor()
]),
mode="train",
keys=tr_keys,
taskname=task),
batch_size=self.config.batch_size,
shuffle=True)
self.val_data_loader = torch.utils.data.DataLoader(
NucleusDataset(self.config.data_root_dir,
train=True,
transform=transforms.Compose([
Normalize(),
Rescale(self.config.patch_size),
ToTensor()
]),
target_transform=transforms.Compose([
Normalize(),
Rescale(self.config.patch_size),
ToTensor()
]),
mode="val",
keys=val_keys,
taskname=self.config.dataset_name),
batch_size=self.config.batch_size,
shuffle=True)
self.test_data_loader = torch.utils.data.DataLoader(
NucleusDataset(self.config.data_root_dir,
train=True,
transform=transforms.Compose([
Normalize(),
Rescale(self.config.patch_size),
ToTensor()
]),
target_transform=transforms.Compose([
Normalize(),
Rescale(self.config.patch_size),
ToTensor()
]),
mode="test",
keys=test_keys,
taskname=self.config.dataset_name),
batch_size=self.config.batch_size,
shuffle=True)
self.model = UNet(num_classes=self.config.num_classes,
in_channels=self.config.in_channels)
#self.model = UNet()
self.model.to(self.device)
self.bce_weight = 0.5
self.optimizer = optim.Adam(self.model.parameters(),
lr=self.config.learning_rate)
self.scheduler = ReduceLROnPlateau(self.optimizer, 'min')
# If directory for checkpoint is provided, we load it.
if self.config.do_load_checkpoint:
if self.config.checkpoint_dir == '':
print(
'checkpoint_dir is empty, please provide directory to load checkpoint.'
)
else:
self.load_checkpoint(name=self.config.checkpoint_dir,
save_types=("model"))
self.save_checkpoint(name="checkpoint_start")
self.elog.print('Experiment set up.')
import torchvision.models as models
from torchvision.utils import make_grid, save_image
from collections import OrderedDict
from utils import visualize_cam, Normalize
from gradcam import GradCAM, GradCAMpp
img_dir = 'examples'
# img_name = 'collies.JPG'
# img_name = 'multiple_dogs.jpg'
# img_name = 'snake.JPEG'
img_name = "real_336_000034.jpg"
img_path = os.path.join(img_dir, img_name)
pil_img = PIL.Image.open(img_path)
normalizer = Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
torch_img = torch.from_numpy(np.asarray(pil_img)).permute(2, 0, 1).unsqueeze(0).float().div(255).cuda()
torch_img = F.upsample(torch_img, size=(224, 224), mode='bilinear', align_corners=False)
normed_torch_img = normalizer(torch_img)
###
# resnet = models.resnet101(pretrained=True)
resnet = models.resnet101() #别忘记传递必要的参数
resnet101_dict = resnet.state_dict()
state_dict = torch.load('./model/2real/extractor_8.pth') #加载预先训练好net-a的.pth文件
new_state_dict = OrderedDict() #不是必要的【from collections import OrderedDict】
new_state_dict = {k:v for k,v in state_dict.items() if k in resnet101_dict} #删除net-b不需要的键
resnet101_dict.update(new_state_dict) #更新参数
noisify3 = Noisify(noise_source3,
snr_high=config['muse_noise_srn_high'],
snr_low=config['muse_noise_srn_low'],
random=True)
rir_files = open(config['rir_path']).read().split('\n')[:-1]
random_rir_reader = RIRSource(rir_files,
random=True,
sample_rate=16000)
reverb = Reverberate(rir_source=random_rir_reader)
muse_augment = RandomChoice([noisify1, noisify2, noisify3])
wav_augments = RandomApply([muse_augment, reverb], 0.25)
transforms += [wav_augments]
melspectrogram = LogMelSpectrogram(**config['fbank'])
transforms += [melspectrogram]
if config['normalize']:
transforms += [Normalize(config['mean_std_file'])]
if config['augment_mel']:
#define spectrogram masking
time_masking = RandomMasking(
max_mask_count=config['max_time_mask'],
max_mask_width=config['max_time_mask_width'],
axis=-1)
freq_masking = RandomMasking(
max_mask_count=config['max_freq_mask'],
max_mask_width=config['max_freq_mask_width'],
axis=-2)
mel_augments = RandomApply([freq_masking, time_masking], p=0.25)
transforms += [mel_augments]
transforms = Compose(transforms)
cfg['scale'] = 0.5
# Original, mean 0.4679, std 0.2699
# Gamma correction: mean 0.3977, std 0.2307
if cfg['scale'] == 0.5:
mnet_v2_mean = [0.4679]
mnet_v2_std = [0.2699]
else:
mnet_v2_mean = [0.4679]
mnet_v2_std = [0.2699]
train_set = OpenEDS(root_path=root_path + 'train',
transform=transforms.Compose([
Rescale(cfg['scale']),
Brightness(brightness=(0.5, 2.75)),
ToTensor(),
Normalize(mnet_v2_mean, mnet_v2_std)
]))
val_set = OpenEDS(root_path=root_path + 'validation',
transform=transforms.Compose([
Rescale(cfg['scale']),
ToTensor(),
Normalize(mnet_v2_mean, mnet_v2_std)
])) #
test_set = OpenEDS(root_path=root_path + 'test',
transform=transforms.Compose([
Rescale(cfg['scale']),
ToTensor(),
Normalize(mnet_v2_mean, mnet_v2_std)
])) #
cal_mean_std_iter = DataLoader(PersianAlphabetDataset(
csv_files=['dataset/train_x.csv', 'dataset/train_y.csv']),
batch_size=args.batch_size)
mean, std = CalMeanStd0(
cal_mean_std_iter) # you have to pass a dataloader object
# ------------------
# --------------------- building dataset
# ------------------
#
print(f"\t✅ building dataset pipeline from CSV files\n")
# normalize image using calculated mean and std per channel
# generally mean and std is a list of per channel values
# in our case one value for std and mean cause we have one channel
# --------------------------------------------------------------------
transform = transforms.Compose([ToTensor(), Normalize(mean=mean, std=std)])
training_transformed = PersianAlphabetDataset(
csv_files=['dataset/train_x.csv', 'dataset/train_y.csv'],
transform=transform)
valid_transformed = PersianAlphabetDataset(
csv_files=['dataset/test_x.csv', 'dataset/test_y.csv'],
transform=transform)
# ---------------------------
# --------------------- building dataloader objects
# ---------------------------
#
print(
f"\t✅ building dataloader objects from training and valid data pipelines\n"
)
# -----------------------------------------------------
def test():
test_path = '/home/WIN-UNI-DUE/smnemada/Master_Thesis/SegThor/data_sub/test'
for patient in tqdm(os.listdir(test_path)):
count = 0
area = 0
file = patient
x = file.split(".")
filename = x[0] + '.' + x[1]
print("patient = ", patient)
test_set = SegThorDataset(test_path,
patient=patient,
phase='test',
transform=transforms.Compose([
Rescale(1.0, labeled=False),
Normalize(labeled=False),
ToTensor(labeled=False)
]))
test_loader = torch.utils.data.DataLoader(dataset=test_set,
batch_size=1,
shuffle=False)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = torch.load("models/model.pt")
model.eval()
'''
with torch.no_grad():
for batch_idx, sample in enumerate(test_loader):
images = sample['image'].to(device, dtype=torch.float)
outputs = model(images)
print("tensor: {} and {} ".format(images.size(), outputs.size()))
images = tensor_to_numpy(images)
max_idx = torch.argmax(outputs, 1, keepdim=True)
max_idx = tensor_to_numpy(max_idx)
print("numpy: {} and {} ".format(images.shape, max_idx.shape))
fig=plt.figure()
fig.add_subplot(1,2,1)
plt.imshow(max_idx)
fig.add_subplot(1,2,2)
plt.imshow(images)
plt.show()
# fig.close()
count = count + 1
if count==150:
break
'''
# '''
seg_vol_2d = zeros([len(test_set), 512, 512])
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = torch.load("models/model.pt")
model.eval()
model.to(device)
with torch.no_grad():
for batch_idx, sample in enumerate(test_loader):
images = sample['image'].to(device, dtype=torch.float)
outputs = model(images)
images = tensor_to_numpy(images)
max_idx = torch.argmax(outputs, 1, keepdim=True)
max_idx = tensor_to_numpy(max_idx)
# for k in range(outputs.size(0)):
# print(max_idx.shape)
slice_v = max_idx[:,:]
slice_v = slice_v.astype(float32)
slice_v = ndimage.interpolation.zoom(slice_v, zoom=1, order=0, mode='nearest', prefilter=True)
seg_vol_2d[count,:,:] = slice_v
count = count + 1
segmentation = sitk.GetImageFromArray(seg_vol_2d, isVector=False)
print(segmentation.GetSize())
sitk.WriteImage(sitk.Cast( segmentation, sitk.sitkUInt8 ), filename, True)
# Test RGB-D data loader, transforms and utilities
data_dir = 'sunrgbd/256'
data_dir = 'sunrgbd/256_lite'
# data_transforms = Compose([Resize(224),
# RandomHorizontalFlip(),
# ToTensor(),
# Normalize(MEAN_RGB, STD_RGB, MEAN_DEPTH, STD_DEPTH))])
# data_transforms = Compose([RandomResizedCrop(224),
# ToTensor(),
# Normalize(MEAN_RGB, STD_RGB, MEAN_DEPTH, STD_DEPTH)])
data_transforms = Compose([
CenterCrop(224),
ToTensor(),
Normalize(MEAN_RGB, STD_RGB, MEAN_DEPTH, STD_DEPTH)
])
rgbd_dataset = ImageFolder(os.path.join(data_dir, 'train'),
data_transforms)
data_loader = DataLoader(rgbd_dataset,
batch_size=4,
shuffle=True,
num_workers=4)
class_names = rgbd_dataset.classes
print(class_names)
rgbd_iter = iter(data_loader)
# Get a batch of training data
from tqdm import tqdm
from utils import load_data, Dx_cross_entropy, Normalize, chunks
from run_fat import attack_step
import numpy as np
import time
Dx_losses = {"logistic_regression": 123, "cross_entropy": Dx_cross_entropy}
losses = {
"logistic_regression": 123,
"cross_entropy": lambda x, y: torch.nn.functional.cross_entropy(x, y)
}
# todo: manage the gpu id
# todo: BUG when n_data mod n_workers is non-zero
norm = Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])
# epsilon = 8./255
def sgd_step(f, data, label, sgd_n_steps, sgd_step_size, sgd_mb_size):
opt = optim.SGD(f.parameters(), lr=sgd_step_size)
for i in range(sgd_n_steps):
opt.zero_grad()
index = range(i * sgd_mb_size, (i + 1) * sgd_mb_size)
loss_f = loss(f(data[index]), label[index])
loss_f.backward()
opt.step()
# def attack_step(model, data, label, epsilon, attack_lr=5e-3, mb_size=128):
# delta = torch.zeros_like(data, requires_grad=True)
f.write('\n')
cfg = dict()
cfg['batch_size'] = 64
cfg['scale'] = 0.5
if cfg['scale'] == 0.5:
mnet_v2_mean = [0.4679]
mnet_v2_std = [0.2699]
else:
mnet_v2_mean = [0.4679]
mnet_v2_std = [0.2699]
train_set = OpenEDS(root_path=root_path + 'train',
transform=transforms.Compose(
[Rescale(cfg['scale']), ToTensor(), Normalize(mnet_v2_mean, mnet_v2_std)])) #
val_set = OpenEDS(root_path=root_path + 'validation',
transform=transforms.Compose(
[Rescale(cfg['scale']), ToTensor(), Normalize(mnet_v2_mean, mnet_v2_std)])) #
test_set = OpenEDS(root_path=root_path + 'test',
transform=transforms.Compose(
[Rescale(cfg['scale']), ToTensor(), Normalize(mnet_v2_mean, mnet_v2_std)])) #
loaders = {'train': torch.utils.data.DataLoader(train_set, batch_size=args.batch_size, shuffle=True,
num_workers=args.num_workers),
'val': torch.utils.data.DataLoader(val_set, batch_size=args.batch_size, shuffle=False,
num_workers=args.num_workers, pin_memory=False),
'test': torch.utils.data.DataLoader(test_set, batch_size=args.batch_size, shuffle=False,
num_workers=args.num_workers, pin_memory=False)
}
[T.Resize((256, 256)),
T.CenterCrop((224, 224)),
T.ToTensor()])
dataset = SelectedImagenet(
imagenet_val_dir='data/imagenet/ILSVRC2012_img_val',
selected_images_csv='data/imagenet/selected_imagenet.csv',
transform=trans)
ori_loader = torch.utils.data.DataLoader(dataset,
batch_size=batch_size,
shuffle=False,
num_workers=8,
pin_memory=False)
model = MODEL.resnet.resnet50(
state_dict_dir='attack/imagenet/models/ckpt/resnet50-19c8e357.pth')
model.eval()
model = nn.Sequential(Normalize(), model)
model.to(device)
if target_attack:
label_switch = torch.tensor(
list(range(500, 1000)) + list(range(0, 500))).long()
label_ls = []
for ind, (ori_img, label) in enumerate(ori_loader):
label_ls.append(label)
if target_attack:
label = label_switch[label]
ori_img = ori_img.to(device)
img = ori_img.clone()
m = 0
for i in range(niters):
import scipy
import math
from torch_geometric.data import Data
from utils import Normalize
import torch_geometric.transforms as T
#torch.cuda.set_device(1)
path = os.path.expanduser("./data/")
label_rate = {
'Cora': 0.052,
'Citeseer': 0.036,
'Pubmed': 0.003,
'Computers': 0.015,
'Photo': 0.021
}
Norm = Normalize()
def get_dataset(name):
if name in ['Cora', 'Citeseer', 'Pubmed']:
dataset = Planetoid(path + name, name)
elif name in ['Computers', 'Photo']:
dataset = Amazon(path + name, name)
else:
raise Exception('Unknown dataset.')
return dataset
def parse_index_file(filename):
"""Parse index file."""
index = []
acc = model.evaluate(test_loader=dtest)
print("Iteration: {}, len(dl): {}, len(du): {},"
" len(dh) {}, acc: {} ".format(iteration,
len(dl.sampler.indices),
len(du.sampler.indices),
len(hcs_idx), acc))
if __name__ == "__main__":
dataset_train = Caltech256Dataset(
root_dir="../caltech256/256_ObjectCategories_train",
transform=transforms.Compose(
[SquarifyImage(),
RandomCrop(224),
Normalize(),
ToTensor()]))
dataset_test = Caltech256Dataset(
root_dir="../caltech256/256_ObjectCategories_test",
transform=transforms.Compose(
[SquarifyImage(),
RandomCrop(224),
Normalize(),
ToTensor()]))
# Creating data indices for training and validation splits:
random_seed = 123
validation_split = 0.1 # 10%
shuffling_dataset = True
batch_size = 16
def train(epochs, batch_size, learning_rate):
torch.manual_seed(1234)
train_loader = torch.utils.data.DataLoader(SegThorDataset(
"/home/WIN-UNI-DUE/smnemada/Master_Thesis/SegThor/data/train",
phase='train',
transform=transforms.Compose([Rescale(1.0),
Normalize(),
ToTensor()])),
batch_size=batch_size,
shuffle=True)
'''
# Loading validation data
val_set = SegThorDataset("/home/WIN-UNI-DUE/smnemada/Master_Thesis/SegThor/data_val", phase='val',
transform=transforms.Compose([
Rescale(0.5),
Normalize(),
ToTensor2()
]))
val_loader = torch.utils.data.DataLoader(dataset=val_set,
batch_size=1,
shuffle=False)
'''
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = UNet().to(device)
model.apply(weight_init)
#optimizer = optim.Adam(model.parameters(), lr=learning_rate) #learning rate to 0.001 for initial stage
optimizer = optim.SGD(model.parameters(),
lr=0.01,
momentum=0.9,
weight_decay=0.00001)
#optimizer = adabound.AdaBound(params = model.parameters(), lr = 0.001, final_lr = 0.1)
for epoch in range(epochs):
f = open('train_output.log', 'a')
f.write('Epoch {}/{}\n'.format(epoch + 1, epochs))
f.write('-' * 10)
running_loss = 0.0
running_loss_label = np.zeros(5)
for batch_idx, sample in enumerate(train_loader):
train_data, labels = sample['image'].to(
device,
dtype=torch.float), sample['label'].to(device,
dtype=torch.uint8)
optimizer.zero_grad()
output = model(train_data)
loss_label, loss = dice_loss2(output, labels)
loss.backward()
optimizer.step()
running_loss += loss.item()
for i in range(5):
running_loss_label[i] += loss_label[i]
epoch_loss = running_loss / len(train_loader)
writer.add_scalar('Train/Loss', epoch_loss, epoch)
f.write("\n Total Dice Loss: {:.4f}\n".format(epoch_loss))
epoch_loss_class = np.true_divide(running_loss_label,
len(train_loader))
f.write(
"Dice per class: Background = {:.4f} Eusophagus = {:.4f} Heart = {:.4f} Trachea = {:.4f} Aorta = {:.4f}\n"
.format(epoch_loss_class[0], epoch_loss_class[1],
epoch_loss_class[2], epoch_loss_class[3],
epoch_loss_class[4]))
#f.write("Dice per class: Background = {:.4f} Eusophagus = {:.4f}\n".format(epoch_loss_class[0], epoch_loss_class[1]))
f.close()
if epoch % 4 == 0:
os.makedirs("models", exist_ok=True)
torch.save(model, "models/model.pt")
# export scalar data to JSON for external processing
writer.export_scalars_to_json("./all_scalars.json")
writer.close()
os.makedirs("models", exist_ok=True)
torch.save(model, "models/model.pt")
# set model
model = createModel(opt) # create a new model
model.setup(opt) # set model
# set dataloader
if opt.augment:
print("with data augmentation")
transformList = [
RandomRotation(10),
RandomResizedCrop(),
Resize(opt.loadSize),
ToTensor(),
Normalize([.485, .456, .406], [.229, .224, .225]),
RandomHorizontalFlip(),
]
else:
print("without data augmentation")
transformList = [
Resize(opt.loadSize),
ToTensor(),
Normalize([.485, .456, .406], [.229, .224, .225])
]
transform = Compose(transformList)
supervisedADataset = createDataset([opt.supervisedADataset],
transform=transform,
outputFile=False)[0]
for i in range(len(segthor_dataset)):
sample = segthor_dataset[i]
# print(i, sample['image'].size())
plt.imshow(sample['image'])
plt.show()
if i == 50:
break
'''
# '''
## Loading data for training phase
segthor_dataset = SegThorDataset(
datapath=
"/home/WIN-UNI-DUE/smnemada/Master_Thesis/SegThor/data_sub/train",
phase='train',
transform=transforms.Compose([
Rescale(1.0, labeled=True),
Normalize(labeled=True),
JointTransform2D(crop=(288, 288), p_flip=0.5),
ToTensor(labeled=True)
]))
for i in range(len(segthor_dataset)):
sample = segthor_dataset[i]
print(i, sample['image'].size(), sample['label'].size())
if i == 5:
break
# '''