urbanlc.model.deep_learning.DeepLearningLCC

class urbanlc.model.deep_learning.DeepLearningLCC(architecture: str, model_params: Dict[str, Any], device: Optional[str] = None, seed: Optional[int] = 0, *args: Any, **kwargs: Any)

Deep-learning-based Land Cover Classification (LCC) model.

Methods

build_model(architecture, model_params)	Build the deep learning model and moves it to the selected device.
construct_transform_map()	Construct transformation maps and functions to map between class labels, indices, and names.
denormalize_class(gt)	Denormalize ground-truth labels.
get_metrics(mode)	Get metrics for the specified mode (Train or Val).
infer(img_path[, convert_numpy, cache, ...])	Perform inference using the deep learning model.
load_model(checkpoint_path)	Load a pre-trained model checkpoint.
normalize_class(gt)	Normalize ground-truth labels.
save_model(filename, current_epoch)	Save the current model checkpoint.
set_device(device)	Set the device for training or inference.
setup_trainer(loss_fn_params, ...)	Set up the trainer with loss functions, optimizer, and scheduler.
to(device)	Move the model to a different device.
train(dataloader_params, trainer_params, ...)	Train the deep learning model.
train_one_epoch(step, train_loader, epoch, ...)	Train the model for one epoch.
validate([root, img_glob, gt_glob, ...])	Validate the deep learning model.