Skip to content

noise option added #18

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
luciana-munhos merged 1 commit into from
Jun 19, 2025
Merged

noise option added #18

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 16 additions & 6 deletions interface.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -89,7 +89,9 @@ def __init__(
load_model_path: str = None,
custom_augment_figuratif=None,
custom_augment_abstrait=None,
n_transforms_augmented = 2
n_transforms_augmented = 2,
noise: bool = False,
noise_std: float = 0.1
):

# Profiler
Expand Down Expand Up @@ -151,6 +153,8 @@ def __init__(
self.augmentation = augmentation
self.input_size = input_size
self.padding = padding
self.noise = noise
self.noise_std = noise_std

# Initialize the dataset and dataloader

Expand All @@ -161,23 +165,29 @@ def __init__(
image_input_size=self.input_size,
custom_augment_abstrait=custom_augment_abstrait,
custom_augment_figuratif=custom_augment_figuratif,
n_transforms_augmented=n_transforms_augmented)

n_transforms_augmented=n_transforms_augmented,
noise=self.noise,
noise_std=self.noise_std)

self.dataset_val = PaintingsDataset(self.data_path+'val/',
augment=False,
transform=self.transform,
padding=self.padding,
image_input_size=self.input_size,
custom_augment_abstrait=None,
custom_augment_figuratif=None,)

custom_augment_figuratif=None,
noise=self.noise,
noise_std=self.noise_std)

self.dataset_test = PaintingsDataset(self.data_path+'test/',
augment=False,
transform=self.transform,
padding=self.padding,
image_input_size=self.input_size,
custom_augment_abstrait=None,
custom_augment_figuratif=None,)
custom_augment_figuratif=None,
noise=self.noise,
noise_std=self.noise_std)

n_figurative = self.dataset_train.len_figurative #only for training
n_abstract = self.dataset_train.len_abstract
Expand Down
17 changes: 16 additions & 1 deletion paintings_dataset.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -73,7 +73,7 @@ class PaintingsDataset(Dataset):
In case of `transform` being True, the dataset will apply transformations to the images or
a custom transformation can be passed as a parameter named `custom_transform`(if None we will use custom).
"""
def __init__(self, data_path, augment= False, transform=False, custom_augment_figuratif=None,custom_augment_abstrait=None, padding: PaddingOptions = PaddingOptions.ZERO, image_input_size: int = 224,n_transforms_augmented=2):
def __init__(self, data_path, augment= False, transform=False, custom_augment_figuratif=None,custom_augment_abstrait=None, padding: PaddingOptions = PaddingOptions.ZERO, image_input_size: int = 224,n_transforms_augmented=2, noise: bool = False, noise_std: float = 0.1):

# Path to the data directory
self.data_path = data_path
Expand Down Expand Up @@ -106,6 +106,10 @@ def __init__(self, data_path, augment= False, transform=False, custom_augment_fi
self.padding = padding
self.image_input_size = image_input_size

# Noise configuration
self.noise = noise
self.noise_std = noise_std


def __len__(self):
return self.total_length
Expand Down Expand Up @@ -176,6 +180,17 @@ def __getitem__(self, idx: int)->Dict[str, Union[torch.Tensor, int]]:
size=(max(imh, int(self.image_input_size / 2 + 1)), max(imw, int(self.image_input_size/2 + 1))),
mode='nearest').squeeze(0).clone()
output['image'] = output['image'].float() / 255.0 # Normalize the image to [0, 1]

# Noise configuration
if self.noise:
noise = torch.randn_like(output['image']) * self.noise_std
output['image'] = torch.clamp(output['image'] + noise, 0.0, 1.0)

# Apply noise to transformed image (if it exists and noise enabled)
if self.noise and self.transform:
noise_transformed = torch.randn_like(output['transformed_image']) * self.noise_std
output['transformed_image'] = torch.clamp(output['transformed_image'] + noise_transformed, 0.0, 1.0)

return output

def change_padding(self, padding: Literal['zero', 'mirror', 'replicate'], image_input_size: int = 224):
Expand Down