def load_unlabeled_data(tar, data_dir='dataset/'):
from torchvision import transforms
folder_tar = data_dir + tar + '/images'
transform = {
'weak':
transforms.Compose([
transforms.Resize([224, 224]),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
]),
'strong':
transforms.Compose([
transforms.Resize([256, 256]),
transforms.RandomCrop(224),
transforms.RandomHorizontalFlip(),
RandAugment(
), # Paper: RandAugment: Practical data augmentation with no separate search
transforms.ToTensor(),
# Cutout(size=16), # (https://github.com/uoguelph-mlrg/Cutout/blob/master/util/cutout.py)
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
])
}
unlabeled_weak_data = datasets.ImageFolder(root=folder_tar,
transform=transform['weak'])
unlabeled_strong_data = datasets.ImageFolder(root=folder_tar,
transform=transform['strong'])
return unlabeled_weak_data, unlabeled_strong_data
def remove_background(input_image):
# Disable ssl verification (This was used to download the trained dataset)
if (not os.environ.get('PYTHONHTTPSVERIFY', '')
and getattr(ssl, '_create_unverified_context', None)):
ssl._create_default_https_context = ssl._create_unverified_context
# Defining the convolutional network with a pretrained dataset that will be used to detect the person in the image
fully_convolutional_network = models.segmentation.fcn_resnet101(
pretrained=True).eval()
# Preprocessing the image and making it become a tensor so it can be used in the convolutional network
image_to_tensor_transform = cvtransforms.Compose([
cvtransforms.ToTensor(),
cvtransforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
])
input_tensor = image_to_tensor_transform(input_image).unsqueeze(0)
# Passing the input structure through the net to get the ouput model with the person classified
output_model = fully_convolutional_network(input_tensor)['out']
# Transforming the tensor into n labeled image
labeled_image = torch.argmax(output_model.squeeze(),
dim=0).detach().cpu().numpy()
# Generating the segmentation mask to remove the background
mask = segmentation_mask(labeled_image)
# Removing the background leaving only the person
result_image = cv2.bitwise_and(input_image,
input_image,
mask=mask.astype(np.uint8))
return result_image
def get_normalization_layer(mean: list, std: list, num_images: int = 1, mode: str = '2d'):
"""Get Z-scoring layer from config
If RGB frames are stacked into tensor N, num_rgb*3, H, W, we need to repeat the mean and std num_rgb times
"""
# if mode == '2d':
mean = mean.copy() * num_images
std = std.copy() * num_images
return transforms.Normalize(mean=mean, std=std)
def get_image_from_url(url=''):
img_transforms = opencv_transform.Compose([
opencv_transform.Resize(int(math.ceil(160 / 0.875)),
interpolation=cv2.INTER_LINEAR),
opencv_transform.CenterCrop(160),
opencv_transform.ToTensor(),
opencv_transform.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
])
# download the image from web if uri is valid.
if (validators.url(url)):
img = Image.open(requests.get(url, stream=True).raw)
else:
img = Image.open(url)
img = img_transforms(img)
return img
def load_data(src, tar, data_dir='dataset', use_cv2=False):
folder_src = os.path.join(os.path.join(data_dir, src), 'images')
folder_tar = os.path.join(os.path.join(data_dir, tar), 'images')
if use_cv2:
import cv2
from opencv_transforms import transforms
def loader_opencv(path: str) -> np.ndarray:
return cv2.imread(path)
transform = {
'train':
transforms.Compose([
transforms.Resize((256, 256), interpolation=cv2.INTER_LINEAR),
transforms.RandomCrop(224),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
]),
'test':
transforms.Compose([
transforms.Resize((224, 224), interpolation=cv2.INTER_LINEAR),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
])
}
source_data = datasets.ImageFolder(root=folder_src,
transform=transform['train'],
loader=loader_opencv)
# source_data_loader = torch.utils.data.DataLoader(source_data, batch_size=batch_size, shuffle=True, **kwargs, drop_last = True)
target_train_data = datasets.ImageFolder(root=folder_tar,
transform=transform['train'],
loader=loader_opencv)
# target_train_loader = torch.utils.data.DataLoader(target_train_data, batch_size=batch_size, shuffle=True, **kwargs, drop_last = True)
target_test_data = datasets.ImageFolder(root=folder_tar,
transform=transform['test'],
loader=loader_opencv)
# target_test_loader = torch.utils.data.DataLoader(target_test_data, batch_size=batch_size, shuffle=True, **kwargs, drop_last = False)
else:
from torchvision import transforms
transform = {
'train':
transforms.Compose([
transforms.Resize((256, 256)),
transforms.RandomCrop(224),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
]),
'test':
transforms.Compose([
transforms.Resize((224, 224)),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
])
}
source_data = datasets.ImageFolder(root=folder_src,
transform=transform['train'])
# source_data_loader = torch.utils.data.DataLoader(source_data, batch_size=batch_size, shuffle=True, **kwargs, drop_last = True)
target_train_data = datasets.ImageFolder(root=folder_tar,
transform=transform['train'])
# target_train_loader = torch.utils.data.DataLoader(target_train_data, batch_size=batch_size, shuffle=True, **kwargs, drop_last = True)
target_test_data = datasets.ImageFolder(root=folder_tar,
transform=transform['test'])
# target_test_loader = torch.utils.data.DataLoader(target_test_data, batch_size=batch_size, shuffle=True, **kwargs, drop_last = False)
return source_data, target_train_data, target_test_data
# Disable ssl verification (This was used to download the trained dataset)
if (not os.environ.get('PYTHONHTTPSVERIFY', '')
and getattr(ssl, '_create_unverified_context', None)):
ssl._create_default_https_context = ssl._create_unverified_context
# Defining the convolutional network with a pretrained dataset that will be used to detect the person in the image
fully_convolutional_network = models.segmentation.fcn_resnet101(
pretrained=True).eval()
input_image = cv2.imread('test_images/input.jpg')
# Preprocessing the image and making it become a tensor so it can be used in the convolutional network
image_to_tensor_transform = cvtransforms.Compose([
cvtransforms.ToTensor(),
cvtransforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
])
input_tensor = image_to_tensor_transform(input_image).unsqueeze(0)
# Passing the input structure through the net to get the ouput model with the person classified
output_model = fully_convolutional_network(input_tensor)['out']
# Transforming the tensor into n labeled image
labeled_image = torch.argmax(output_model.squeeze(),
dim=0).detach().cpu().numpy()
# Generating the segmentation mask to remove the background
mask = segmentation_mask(labeled_image)
# Removing the background leaving only the person