Read Zarr outputs¶

Zarr output is written in batches: one folder per batch, each holding an images.zarr archive and an images_metadata.parquet file. The layout is described in the output formats reference.

Find a source's cutout¶

The Zarr archives store no source IDs. Row i of the metadata describes image i of the archive in the same folder, so look up a SourceID by its row position:

import pandas as pd

batch = "output_path/batch_cutout_process_000_abc123"
metadata = pd.read_parquet(f"{batch}/images_metadata.parquet")
index = metadata.index[metadata["source_id"] == "TILE_102018666_12345"][0]

Load images¶

import zarr

archive = zarr.open(f"{batch}/images.zarr", mode="r")
images = archive["images"]  # shape (n_images, H, W, C)
cutout = images[index]

Plot a sample¶

import numpy as np
from matplotlib import pyplot as plt

fig, axes = plt.subplots(4, 4, figsize=(8, 8))
for ax in axes.flatten():
    ax.imshow(images[np.random.randint(0, images.shape[0])], cmap="gray", origin="lower")
    ax.axis("off")
plt.tight_layout()
plt.show()

With FITS output, each source gets its own file, and the information is in the header of the PRIMARY extension.