Datasets:
File size: 5,868 Bytes
1788f61 | 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 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 | #!/usr/bin/env python3
"""Create public metadata and split indexes from the canonical CARLA tree."""
from __future__ import annotations
import argparse
import json
import re
import shutil
from datetime import datetime, timezone
from pathlib import Path
WEATHERS = ("ClearDay", "ClearNight", "HeavyFoggyNight", "HeavyRainFoggyNight")
MODALITIES = ("image_2", "gt_image_2", "depth_u16", "depth_meters", "normal", "calib")
STEM_RE = re.compile(r"^(Town04|Town05|Town06)_(ClearDay|ClearNight|HeavyFoggyNight|HeavyRainFoggyNight)_(\d{6})$")
def sanitize_manifest(source: Path, target: Path) -> None:
data = json.loads(source.read_text(encoding="utf-8"))
# Do not publish workstation-specific absolute paths.
if isinstance(data, dict):
if "source" in data:
data["source"] = "CARLA_Unified_Dataset (local source; not redistributed)"
if "output" in data:
data["output"] = "."
target.write_text(json.dumps(data, indent=2) + "\n", encoding="utf-8")
def main() -> int:
ap = argparse.ArgumentParser()
ap.add_argument("--source-root", type=Path, required=True)
ap.add_argument("--release-root", type=Path, required=True)
args = ap.parse_args()
src = args.source_root.resolve()
out = args.release_root.resolve()
(out / "metadata").mkdir(parents=True, exist_ok=True)
(out / "splits").mkdir(parents=True, exist_ok=True)
sanitize_manifest(src / "MANIFEST_reconciled_town_split.json", out / "metadata/MANIFEST_reconciled_town_split.json")
shutil.copy2(src / "02_reconciled_protocol.json", out / "metadata/02_reconciled_protocol.json")
shutil.copy2(src / "SELECTED_STEMS.sha256", out / "metadata/SELECTED_STEMS.sha256")
split_records = {}
for split in ("training", "validation"):
stems = sorted(p.stem for p in (src / split / "image_2").glob("*.png"))
records = []
for stem in stems:
match = STEM_RE.match(stem)
if not match:
raise ValueError(f"invalid stem: {stem}")
town, weather, frame = match.groups()
records.append((stem, town, weather, int(frame)))
split_records[split] = records
(out / "splits" / f"{split}.txt").write_text(
"".join(f"{r[0]}\n" for r in records), encoding="utf-8"
)
(out / "splits" / f"{split}.tsv").write_text(
"stem\ttown\tweather\tframe\n" + "".join(
f"{stem}\t{town}\t{weather}\t{frame:06d}\n" for stem, town, weather, frame in records
), encoding="utf-8"
)
source_manifest = json.loads((src / "MANIFEST_reconciled_town_split.json").read_text(encoding="utf-8"))
protocol = json.loads((src / "02_reconciled_protocol.json").read_text(encoding="utf-8"))
archives = {}
archive_manifest = out / "archive_manifest.json"
if archive_manifest.exists():
archives = json.loads(archive_manifest.read_text(encoding="utf-8")).get("archives", [])
manifest = {
"dataset": "CARLA-MWRS",
"version": "1.0.0",
"status": "PASS",
"release_date": datetime.now(timezone.utc).date().isoformat(),
"protocol": protocol.get("protocol", "CARLA-MWRS Town05+Town06 training / Town04 held-out validation"),
"selection_seed": source_manifest.get("seed", protocol.get("seed", 42)),
"weathers": list(WEATHERS),
"model_input_size_hw": protocol.get("input_size", [512, 1024]),
"materialized_file_size_hw": [384, 1248],
"splits": {
split: {
"samples": len(records),
"towns": sorted({r[1] for r in records}),
"weather_counts": {w: sum(r[2] == w for r in records) for w in WEATHERS},
"stem_sha256": source_manifest.get("filename_sha256", {}).get(split),
"archives": [f"{split}/{m}.tar.zst" for m in MODALITIES],
}
for split, records in split_records.items()
},
"modalities": {
"image_2": {"extension": ".png", "shape": [384, 1248, 3], "dtype": "uint8", "encoding": "RGB PNG"},
"gt_image_2": {"extension": ".png", "shape": [384, 1248], "dtype": "uint8", "values": [0, 255], "encoding": "binary road mask"},
"depth_u16": {"extension": ".png", "shape": [384, 1248], "dtype": "uint16", "unit": "millimetres", "saturation": 65535},
"depth_meters": {"extension": ".npy", "shape": [384, 1248], "dtype": "float32", "unit": "metres"},
"normal": {"extension": ".npy", "shape": [3, 384, 1248], "dtype": "float32", "layout": "C,H,W", "unit": "unit camera-frame vector"},
"calib": {"extension": ".txt", "records": ["P2: 12 values (3x4)", "Vehicle_pos: 3 values"]},
},
"depth_u16_conversion": "clip(floor(float32(depth_meters) * float32(1000)), 0, 65535).astype(uint16)",
"archive_layout": "one deterministic tar.zst per split and modality; archive members are <modality>/<stem><extension>",
"archives": archives,
"source_manifests": [
"metadata/MANIFEST_reconciled_town_split.json",
"metadata/02_reconciled_protocol.json",
"metadata/SELECTED_STEMS.sha256",
],
"validation": {
"validator": "scripts/validate_release.py",
"source_validation_report": "source_validation_report.json",
"source_file_hashes": "SOURCE_FILES.sha256",
"archive_validator": "scripts/validate_archives.py",
"archive_validation_report": "archive_validation_report.json",
},
}
(out / "dataset_manifest.json").write_text(json.dumps(manifest, indent=2, sort_keys=True) + "\n", encoding="utf-8")
print(f"wrote metadata for {sum(len(v) for v in split_records.values())} samples to {out}")
return 0
if __name__ == "__main__":
raise SystemExit(main())
|