LKCell / cell_segmentation /datasets /dataset_coordinator.py
qingke1's picture
initial commit
aea73e2
# -*- coding: utf-8 -*-
# Coordinate the datasets, used to select the right dataset with corresponding setting
#
# @ Fabian Hörst, fabian.hoerst@uk-essen.de
# Institute for Artifical Intelligence in Medicine,
# University Medicine Essen
from typing import Callable
from torch.utils.data import Dataset
from cell_segmentation.datasets.conic import CoNicDataset
from cell_segmentation.datasets.pannuke import PanNukeDataset
def select_dataset(
dataset_name: str, split: str, dataset_config: dict, transforms: Callable = None
) -> Dataset:
"""Select a cell segmentation dataset from the provided ones, currently just PanNuke is implemented here
Args:
dataset_name (str): Name of dataset to use.
Must be one of: [pannuke, lizzard]
split (str): Split to use.
Must be one of: ["train", "val", "validation", "test"]
dataset_config (dict): Dictionary with dataset configuration settings
transforms (Callable, optional): PyTorch Image and Mask transformations. Defaults to None.
Raises:
NotImplementedError: Unknown dataset
Returns:
Dataset: Cell segmentation dataset
"""
assert split.lower() in [
"train",
"val",
"validation",
"test",
], "Unknown split type!"
if dataset_name.lower() == "pannuke":
if split == "train":
folds = dataset_config["train_folds"]
if split == "val" or split == "validation":
folds = dataset_config["val_folds"]
if split == "test":
folds = dataset_config["test_folds"]
dataset = PanNukeDataset(
dataset_path=dataset_config["dataset_path"],
folds=folds,
transforms=transforms,
stardist=dataset_config.get("stardist", False),
regression=dataset_config.get("regression_loss", False),
)
elif dataset_name.lower() == "conic":
if split == "train":
folds = dataset_config["train_folds"]
if split == "val" or split == "validation":
folds = dataset_config["val_folds"]
if split == "test":
folds = dataset_config["test_folds"]
dataset = CoNicDataset(
dataset_path=dataset_config["dataset_path"],
folds=folds,
transforms=transforms,
stardist=dataset_config.get("stardist", False),
regression=dataset_config.get("regression_loss", False),
# TODO: Stardist and regression loss
)
else:
raise NotImplementedError(f"Unknown dataset: {dataset_name}")
return dataset