File size: 2,519 Bytes
0832232
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
from __future__ import annotations

import argparse
import gzip
import hashlib
import json
import struct
from pathlib import Path

import numpy as np


def main() -> int:
    parser = argparse.ArgumentParser(description="Export UFO-MNIST NPZ arrays to MNIST-style IDX gzip files.")
    parser.add_argument("--dataset", type=Path, default=Path("data/ufo_mnist_v1/ufo_mnist_28x28.npz"))
    parser.add_argument("--output", type=Path, default=Path("data/ufo"))
    parser.add_argument("--checksums", type=Path, default=Path("data/ufo_mnist_v1/checksums.json"))
    args = parser.parse_args()

    data = np.load(args.dataset)
    args.output.mkdir(parents=True, exist_ok=True)

    files = {
        "train-images-idx3-ubyte.gz": _write_images(args.output / "train-images-idx3-ubyte.gz", data["train_images"]),
        "train-labels-idx1-ubyte.gz": _write_labels(args.output / "train-labels-idx1-ubyte.gz", data["train_labels"]),
        "t10k-images-idx3-ubyte.gz": _write_images(args.output / "t10k-images-idx3-ubyte.gz", data["test_images"]),
        "t10k-labels-idx1-ubyte.gz": _write_labels(args.output / "t10k-labels-idx1-ubyte.gz", data["test_labels"]),
    }

    if args.checksums.exists():
        checksums = json.loads(args.checksums.read_text(encoding="utf-8"))
    else:
        checksums = {}
    for name, digest in files.items():
        checksums[f"data/ufo/{name}"] = digest
    args.checksums.write_text(json.dumps(checksums, indent=2, sort_keys=True) + "\n", encoding="utf-8")

    for name, digest in files.items():
        print(f"{name} {digest}")
    return 0


def _write_images(path: Path, images: np.ndarray) -> str:
    images = np.asarray(images, dtype=np.uint8)
    if images.ndim != 3:
        raise ValueError("images must be shaped (N, rows, cols)")
    header = struct.pack(">IIII", 2051, images.shape[0], images.shape[1], images.shape[2])
    return _write_gzip(path, header + images.tobytes())


def _write_labels(path: Path, labels: np.ndarray) -> str:
    labels = np.asarray(labels, dtype=np.uint8)
    if labels.ndim != 1:
        raise ValueError("labels must be shaped (N,)")
    header = struct.pack(">II", 2049, labels.shape[0])
    return _write_gzip(path, header + labels.tobytes())


def _write_gzip(path: Path, payload: bytes) -> str:
    with gzip.GzipFile(filename="", mode="wb", fileobj=path.open("wb"), mtime=0) as handle:
        handle.write(payload)
    return hashlib.sha256(path.read_bytes()).hexdigest()


if __name__ == "__main__":
    raise SystemExit(main())