scripts/view_ct.py

#!/usr/bin/env python3
"""
Napari viewer for CT data from ORNL LPBF Cylinders dataset.

Usage:
    uv run scripts/view_ct.py [--source PATH]

Loads the X-ray CT data and flaw segmentation lazily using dask,
then opens an interactive napari viewer.
"""

import argparse
from pathlib import Path

import dask.array as da
import h5py
import napari


def load_ct_lazy(hdf5_path: Path) -> tuple[da.Array, da.Array, h5py.File]:
    """Load CT datasets lazily using dask.

    Returns dask arrays that only load data on demand, plus the open file handle.
    """
    f = h5py.File(hdf5_path, "r")

    ct_data = f["slices/registered_data/x-ray_ct_data"]
    ct_flaw = f["slices/registered_data/x-ray_ct_flaw"]

    # Wrap as dask arrays for lazy loading
    # Use chunks aligned with the data for efficient access
    ct_data_dask = da.from_array(ct_data, chunks=(1, 1024, 1024))
    ct_flaw_dask = da.from_array(ct_flaw, chunks=(1, 1024, 1024))

    return ct_data_dask, ct_flaw_dask, f


def main():
    parser = argparse.ArgumentParser(description="View CT data with napari")
    parser.add_argument(
        "--source",
        type=Path,
        default=Path("source/2024-05-01 M2 AMMTO Fatigue Blanks 05.hdf5"),
        help="Path to HDF5 file",
    )
    args = parser.parse_args()

    if not args.source.exists():
        print(f"Error: {args.source} not found")
        return 1

    print(f"Loading CT data from {args.source}...")
    print("  (Data loads lazily - initial startup is fast)")

    ct_data, ct_flaw, file_handle = load_ct_lazy(args.source)

    print(f"  CT data shape: {ct_data.shape}")
    print(f"  CT flaw shape: {ct_flaw.shape}")
    print()
    print("Opening napari viewer...")
    print("  - Scroll mouse wheel to move through layers")
    print("  - Use layer controls to toggle CT/flaw visibility")
    print("  - Adjust contrast with the layer controls")

    viewer = napari.Viewer(title="ORNL LPBF CT Viewer")

    # Add raw CT data
    viewer.add_image(
        ct_data,
        name="X-ray CT",
        colormap="gray",
        contrast_limits=[0, 65535],
        blending="translucent",
    )

    # Add flaw segmentation as labels
    viewer.add_labels(
        ct_flaw,
        name="Flaw Segmentation",
        opacity=0.5,
    )

    # Set initial view to middle slice
    viewer.dims.current_step = (ct_data.shape[0] // 2, 0, 0)

    napari.run()

    # Clean up
    file_handle.close()
    return 0


if __name__ == "__main__":
    exit(main())