| |
| """Smoke-test training-ready NPZ manifests with a PyTorch DataLoader.""" |
|
|
| from __future__ import annotations |
|
|
| import argparse |
| from pathlib import Path |
|
|
| import numpy as np |
| import pandas as pd |
| import torch |
| from torch.utils.data import DataLoader, Dataset |
|
|
|
|
| def center_crop_or_pad(array: np.ndarray, target_shape: tuple[int, ...]) -> np.ndarray: |
| result = np.zeros(target_shape, dtype=array.dtype) |
| src_slices = [] |
| dst_slices = [] |
| for src_size, dst_size in zip(array.shape, target_shape): |
| if src_size >= dst_size: |
| start = (src_size - dst_size) // 2 |
| src_slices.append(slice(start, start + dst_size)) |
| dst_slices.append(slice(0, dst_size)) |
| else: |
| start = (dst_size - src_size) // 2 |
| src_slices.append(slice(0, src_size)) |
| dst_slices.append(slice(start, start + src_size)) |
| result[tuple(dst_slices)] = array[tuple(src_slices)] |
| return result |
|
|
|
|
| class WawTaceDataset(Dataset): |
| def __init__(self, manifest_path: Path, patch_size: tuple[int, int, int]): |
| self.root = manifest_path.resolve().parents[1] |
| self.df = pd.read_csv(manifest_path) |
| self.patch_size = patch_size |
|
|
| def __len__(self) -> int: |
| return len(self.df) |
|
|
| def __getitem__(self, index: int) -> dict[str, torch.Tensor]: |
| row = self.df.iloc[index] |
| with np.load(self.root / row["npz_path"]) as data: |
| image = center_crop_or_pad(data["image"], (4, *self.patch_size)) |
| liver = center_crop_or_pad(data["liver_mask"], self.patch_size) |
| tumor = center_crop_or_pad(data["tumor_mask"], self.patch_size) |
| phase_available = data["phase_available"].astype(np.float32) |
| label_progression = np.asarray(row["label_progression"], dtype=np.float32) |
| return { |
| "image": torch.from_numpy(image.astype(np.float32)), |
| "liver_mask": torch.from_numpy(liver.astype(np.int64)), |
| "tumor_mask": torch.from_numpy(tumor.astype(np.int64)), |
| "phase_available": torch.from_numpy(phase_available), |
| "label_progression": torch.from_numpy(label_progression[None]), |
| } |
|
|
|
|
| class MsdLiverDataset(Dataset): |
| def __init__(self, manifest_path: Path, patch_size: tuple[int, int, int]): |
| self.root = manifest_path.resolve().parents[1] |
| self.df = pd.read_csv(manifest_path) |
| self.patch_size = patch_size |
|
|
| def __len__(self) -> int: |
| return len(self.df) |
|
|
| def __getitem__(self, index: int) -> dict[str, torch.Tensor]: |
| row = self.df.iloc[index] |
| with np.load(self.root / row["npz_path"]) as data: |
| image = center_crop_or_pad(data["image"], (1, *self.patch_size)) |
| label = center_crop_or_pad(data["label"], self.patch_size) |
| return { |
| "image": torch.from_numpy(image.astype(np.float32)), |
| "label": torch.from_numpy(label.astype(np.int64)), |
| } |
|
|
|
|
| class HccTaceSegDataset(MsdLiverDataset): |
| pass |
|
|
|
|
| def parse_patch_size(value: str) -> tuple[int, int, int]: |
| parts = tuple(int(x) for x in value.split(",")) |
| if len(parts) != 3: |
| raise ValueError("--patch-size must be z,y,x") |
| return parts |
|
|
|
|
| def main() -> None: |
| parser = argparse.ArgumentParser(description=__doc__) |
| parser.add_argument("--dataset", choices=["waw", "msd", "hcc"], default="waw") |
| parser.add_argument("--batch-size", type=int, default=2) |
| parser.add_argument("--patch-size", default="96,128,128", help="Patch size in z,y,x order.") |
| args = parser.parse_args() |
|
|
| patch_size = parse_patch_size(args.patch_size) |
| if args.dataset == "waw": |
| dataset: Dataset = WawTaceDataset(Path("manifests/waw_tace_training_manifest.csv"), patch_size) |
| elif args.dataset == "msd": |
| dataset = MsdLiverDataset(Path("manifests/msd_liver_training_manifest.csv"), patch_size) |
| else: |
| dataset = HccTaceSegDataset(Path("manifests/hcc_tace_seg_training_manifest.csv"), patch_size) |
|
|
| loader = DataLoader(dataset, batch_size=args.batch_size, shuffle=True, num_workers=0) |
| batch = next(iter(loader)) |
| for key, value in batch.items(): |
| print(key, tuple(value.shape), value.dtype) |
|
|
|
|
| if __name__ == "__main__": |
| main() |
|
|
