File size: 6,309 Bytes
edbf290 8b5a68d 62b7002 fc49781 62b7002 8b5a68d 62b7002 e9802f8 8b5a68d fc49781 8b5a68d fc49781 8b5a68d fc49781 8b5a68d 62b7002 8b5a68d 62b7002 8b5a68d fc49781 8b5a68d 62b7002 8b5a68d fc49781 8b5a68d | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 | """Run the 2D floor-plan component detector from Hugging Face Hub or local files.
Example:
python inference.py plan.png --output-dir results/plan
"""
from __future__ import annotations
import argparse
import json
from pathlib import Path
import numpy as np
import torch
from huggingface_hub import hf_hub_download
from PIL import Image, ImageDraw
from rfdetr import RFDETRMedium
DEFAULT_REPO_ID = "OsamaMo/2dplan2strct"
COLORS = {
"wall": "#e63946",
"room": "#457b9d",
"door": "#f4a261",
"window": "#2a9d8f",
}
class FloorPlanDetector:
"""Load RF-DETR once, then run it on one or more floor-plan images."""
def __init__(
self,
repo_id: str = DEFAULT_REPO_ID,
*,
revision: str | None = None,
model_dir: Path | None = None,
device: str = "auto",
) -> None:
self.repo_id = repo_id
if device not in {"auto", "cpu", "cuda"}:
raise ValueError("device must be 'auto', 'cpu', or 'cuda'")
self.device = "cuda" if device == "auto" and torch.cuda.is_available() else device
if self.device == "auto":
self.device = "cpu"
if self.device == "cuda" and not torch.cuda.is_available():
raise RuntimeError("CUDA was requested, but no CUDA GPU is available")
if model_dir is None:
config_path = Path(hf_hub_download(repo_id=repo_id, filename="config.json", revision=revision))
config = json.loads(config_path.read_text(encoding="utf-8"))
weights = hf_hub_download(repo_id=repo_id, filename=config["checkpoint"], revision=revision)
else:
model_dir = Path(model_dir)
config = json.loads((model_dir / "config.json").read_text(encoding="utf-8"))
weights = str(model_dir / config["checkpoint"])
if config["variant"] != "RFDETRMedium":
raise ValueError(f"Unsupported model variant: {config['variant']}")
self.class_names = config["class_names"]
self.model = RFDETRMedium(
pretrain_weights=weights,
resolution=config["resolution"],
num_classes=config["num_classes"],
device=self.device,
)
if self.model.class_names and self.model.class_names != self.class_names:
raise ValueError("Checkpoint class names do not match config.json")
def predict(self, image_path: str | Path, *, threshold: float = 0.35) -> dict:
"""Return boxes in original-image pixel coordinates, sorted by confidence."""
if not 0 <= threshold <= 1:
raise ValueError("threshold must be between 0 and 1")
image_path = Path(image_path)
with Image.open(image_path) as source:
image = np.array(source.convert("RGB"))
height, width = image.shape[:2]
result = self.model.predict(image, threshold=threshold)
detections = [
{
"label": self.class_names[int(class_id)],
"score": float(score),
"box_xyxy": [float(value) for value in box],
}
for box, score, class_id in zip(result.xyxy, result.confidence, result.class_id)
]
detections.sort(key=lambda item: item["score"], reverse=True)
return {
"model": self.repo_id,
"image": str(image_path),
"width": width,
"height": height,
"threshold": threshold,
"detections": detections,
}
def save_prediction(prediction: dict, image_path: str | Path, output_dir: str | Path) -> tuple[Path, Path]:
"""Write a JSON result and an annotated PNG without changing the source image."""
output_dir = Path(output_dir)
output_dir.mkdir(parents=True, exist_ok=True)
json_path = output_dir / "predictions.json"
image_out = output_dir / "annotated.png"
json_path.write_text(json.dumps(prediction, indent=2) + "\n", encoding="utf-8")
with Image.open(image_path) as source:
image = source.convert("RGB")
draw = ImageDraw.Draw(image)
line_width = max(2, min(image.size) // 300)
for item in reversed(prediction["detections"]):
color = COLORS.get(item["label"], "#ffffff")
box = item["box_xyxy"]
draw.rectangle(box, outline=color, width=line_width)
label = f"{item['label']} {item['score']:.2f}"
text_box = draw.textbbox((box[0], box[1]), label)
text_height = text_box[3] - text_box[1]
label_y = max(0, box[1] - text_height - 4)
draw.rectangle((box[0], label_y, box[0] + text_box[2] - text_box[0] + 4, label_y + text_height + 4), fill=color)
draw.text((box[0] + 2, label_y + 2), label, fill="white")
image.save(image_out)
return json_path, image_out
def main() -> None:
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("image", type=Path, help="PNG or JPEG floor-plan image")
parser.add_argument("--repo-id", default=DEFAULT_REPO_ID, help="Hugging Face model repository")
parser.add_argument("--revision", help="Optional Hub commit SHA or tag for reproducible downloads")
parser.add_argument("--model-dir", type=Path, help="Use previously downloaded config and weights offline")
parser.add_argument("--device", choices=("auto", "cpu", "cuda"), default="auto",
help="Run on CPU or CUDA; auto uses CUDA when available")
parser.add_argument("--threshold", type=float, default=0.35, help="Confidence threshold from 0 to 1")
parser.add_argument("--output-dir", type=Path, default=Path("output/prediction"))
args = parser.parse_args()
if not args.image.is_file():
parser.error(f"image does not exist: {args.image}")
if not 0 <= args.threshold <= 1:
parser.error("--threshold must be between 0 and 1")
detector = FloorPlanDetector(args.repo_id, revision=args.revision, model_dir=args.model_dir,
device=args.device)
prediction = detector.predict(args.image, threshold=args.threshold)
json_path, image_path = save_prediction(prediction, args.image, args.output_dir)
print(f"{len(prediction['detections'])} detections")
print(f"JSON: {json_path}")
print(f"Annotated image: {image_path}")
if __name__ == "__main__":
main()
|