brails.processors.year_built_classifier.lib.datasets module
Class object that prepares data for the year built classifier.
- class brails.processors.year_built_classifier.lib.datasets.YearBuiltFolder(image_folder, soft_labels=False, gaussian_std=1.5, transforms=None, classes=None, calc_perf=False)
Bases:
Dataset
A PyTorch Dataset class for loading and preprocessing images.
- img_paths
List of image file paths.
- Type:
list[str]
- filenames
List of image file names.
- Type:
list[str]
- labels
List of labels corresponding to images.
- Type:
Union[list[str], list[np.ndarray]]
- calc_perf
Whether to calculate performance metrics.
- Type:
bool
- soft_labels
Whether soft labels are assigned.
- Type:
bool
- classes
Array of unique class labels.
- Type:
np.ndarray
- class_weights
Weights for each class based on their frequencies.
- Type:
dict[str, float]
- train_weights
Weights for individual samples for training.
- Type:
list[float]
- transforms
Image transformations applied during loading.
- Type:
Any
- __len__()
Returns the total number of samples in the dataset.
- __getitem__(index
int) -> tuple[torch.Tensor,torch.Tensor|list],str]: Returns the image, label, and image path for the given index.
- loader(path
str) -> Image.Image: Loads and returns an image from the specified path.
- loader(path)
Load an image from the specified path and converts it to RGB format.
- Parameters:
path (str) – The file path of the image to load.
- Returns:
The loaded image in RGB format.
- Return type:
Image.Image