Source code for sc2bench.models.segmentation.registry

from torchdistill.common.main_util import load_ckpt
from torchdistill.models.official import get_semantic_segmentation_model
from torchdistill.models.registry import get_model, register_model

from sc2bench.models.segmentation.base import check_if_updatable_segmentation_model

SEGMENTATION_MODEL_CLASS_DICT = dict()
SEGMENTATION_MODEL_FUNC_DICT = dict()




def register_segmentation_model_class(cls):
    """
    Registers a semantic segmentation model class.

    :param cls: semantic segmentation model to be registered
    :type cls: class
    :return: registered semantic segmentation model class
    :rtype: class
    """
    SEGMENTATION_MODEL_CLASS_DICT[cls.__name__] = cls
    register_model(cls)
    return cls





def register_segmentation_model_func(func):
    """
    Registers a function to build a semantic segmentation model.

    :param func: function to build a semantic segmentation model to be registered
    :type func: typing.Callable
    :return: registered function
    :rtype: typing.Callable
    """
    SEGMENTATION_MODEL_FUNC_DICT[func.__name__] = func
    register_model(func)
    return func





def get_segmentation_model(cls_or_func_name, **kwargs):
    """
    Gets a semantic segmentation model.

    :param cls_or_func_name: model class or function name
    :type cls_or_func_name: str
    :return: semantic segmentation model
    :rtype: nn.Module or None
    """
    if cls_or_func_name in SEGMENTATION_MODEL_CLASS_DICT:
        return SEGMENTATION_MODEL_CLASS_DICT[cls_or_func_name](**kwargs)
    elif cls_or_func_name in SEGMENTATION_MODEL_FUNC_DICT:
        return SEGMENTATION_MODEL_FUNC_DICT[cls_or_func_name](**kwargs)
    return None





def load_segmentation_model(model_config, device, strict=True):
    """
    Loads a semantic segmentation model.

    :param model_config: model configuration
    :type model_config: dict
    :param device: device
    :type device: torch.device or str
    :param strict: whether to strictly enforce that the keys in state_dict match
            the keys returned by this module's `state_dict()` function.
    :type strict: bool
    :return: semantic segmentation model
    :rtype: nn.Module
    """
    model = get_semantic_segmentation_model(model_config)
    model_name = model_config['key']
    if model is None:
        model = get_segmentation_model(model_name, **model_config['kwargs'])

    if model is None:
        repo_or_dir = model_config.get('repo_or_dir', None)
        model = get_model(model_name, repo_or_dir, **model_config['kwargs'])

    if model_config.get('update_before_ckpt', False) and check_if_updatable_segmentation_model(model):
        model.update()

    src_ckpt_file_path = model_config.get('src_ckpt', None)
    if src_ckpt_file_path is not None:
        load_ckpt(src_ckpt_file_path, model=model, strict=strict)
    return model.to(device)