Buckets:
| """Render review-only DIOR-R bridge source-label contact sheets.""" | |
| from __future__ import annotations | |
| import argparse | |
| import json | |
| from pathlib import Path | |
| import xml.etree.ElementTree as ET | |
| import zipfile | |
| from PIL import Image, ImageDraw | |
| ROOT = Path(__file__).resolve().parents[1] | |
| def polygons(payload: bytes): | |
| root = ET.fromstring(payload) | |
| tags = ("x_left_top", "y_left_top", "x_right_top", "y_right_top", "x_right_bottom", "y_right_bottom", "x_left_bottom", "y_left_bottom") | |
| for obj in root.findall("object"): | |
| box = obj.find("robndbox") | |
| if box is None: | |
| continue | |
| points = [(float(box.findtext(tags[i])), float(box.findtext(tags[i + 1]))) for i in range(0, len(tags), 2)] | |
| yield obj.findtext("name", "").strip().lower(), obj.findtext("difficult", "0").strip() == "1", points | |
| def render(queue: Path, output: Path, columns: int = 4) -> list[Path]: | |
| records = json.loads(queue.read_text())["records"] | |
| output.mkdir(parents=True, exist_ok=True) | |
| rendered = [] | |
| cell, title = 384, 24 | |
| archive = ROOT / records[0]["source_annotation_archive"] | |
| with zipfile.ZipFile(archive) as annotations: | |
| for page_start in range(0, len(records), columns * 2): | |
| page_records = records[page_start : page_start + columns * 2] | |
| rows = (len(page_records) + columns - 1) // columns | |
| page = Image.new("RGB", (columns * cell, rows * (cell + title)), "black") | |
| for index, record in enumerate(page_records): | |
| image = Image.open(ROOT / record["image"]).convert("RGB") | |
| labels = list(polygons(annotations.read(record["source_annotation_member"]))) | |
| bridge_points = [point for name, difficult, points in labels if name == "bridge" and not difficult for point in points] | |
| if not bridge_points: | |
| raise ValueError(f"no usable bridge labels in {record['id']}") | |
| xs, ys = [point[0] for point in bridge_points], [point[1] for point in bridge_points] | |
| left, top = max(0, int(min(xs)) - 160), max(0, int(min(ys)) - 160) | |
| right, bottom = min(image.width, int(max(xs)) + 160), min(image.height, int(max(ys)) + 160) | |
| image = image.crop((left, top, right, bottom)) | |
| draw = ImageDraw.Draw(image) | |
| for name, difficult, points in labels: | |
| points = [(x - left, y - top) for x, y in points] | |
| color = "lime" if name == "bridge" and not difficult else ("orange" if name == "bridge" else "red") | |
| draw.line(points + [points[0]], fill=color, width=3) | |
| image.thumbnail((cell, cell), Image.Resampling.LANCZOS) | |
| x, y = (index % columns) * cell, (index // columns) * (cell + title) | |
| page.paste(image, (x, y + title)) | |
| ImageDraw.Draw(page).text((x + 4, y + 4), record["id"], fill="white") | |
| path = output / f"page-{page_start // (columns * 2) + 1:02d}.jpg" | |
| page.save(path, quality=92) | |
| rendered.append(path) | |
| return rendered | |
| def main() -> None: | |
| parser = argparse.ArgumentParser() | |
| parser.add_argument("--queue", type=Path, required=True) | |
| parser.add_argument("--output", type=Path, required=True) | |
| args = parser.parse_args() | |
| paths = render(args.queue, args.output) | |
| print(json.dumps({"pages": [str(path) for path in paths]})) | |
| if __name__ == "__main__": | |
| main() | |
Xet Storage Details
- Size:
- 3.38 kB
- Xet hash:
- c70dcec6a66e60ac16705cc2d1f581e199ab9ca0d8bf43d227defd42eda8939d
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.