Initial commit: 包装审核 POC、Docker 与前后端

Made-with: Cursor

scripts/detect_layout_boxes.py

@@ -0,0 +1,108 @@
+#!/usr/bin/env python3
+from __future__ import annotations
+
+import argparse
+import json
+import shutil
+import subprocess
+from pathlib import Path
+
+import cv2
+
+from backend.app.layout_cv import process_image
+
+
+def detect_kind(path: Path) -> str:
+    suffix = path.suffix.lower()
+    if suffix in {".jpg", ".jpeg", ".png", ".webp", ".bmp", ".tif", ".tiff"}:
+        return "image"
+    if suffix == ".pdf":
+        return "pdf"
+    if suffix == ".ai":
+        with path.open("rb") as file:
+            head = file.read(4096)
+        if b"%PDF-" in head or b"%%PDF" in head:
+            return "pdf-compatible-ai"
+        return "eps-like-ai"
+    raise ValueError(f"Unsupported input type: {path.suffix}")
+
+
+def ensure_raster_image(source: Path, workdir: Path) -> Path:
+    kind = detect_kind(source)
+    if kind == "image":
+        return source
+
+    pdf_path = source
+    if kind in {"pdf-compatible-ai", "eps-like-ai"}:
+        converted_dir = workdir / "converted"
+        converted_dir.mkdir(parents=True, exist_ok=True)
+        command = [
+            "soffice",
+            "--headless",
+            "--convert-to",
+            "pdf",
+            "--outdir",
+            str(converted_dir),
+            str(source),
+        ]
+        completed = subprocess.run(command, capture_output=True, text=True, check=False)
+        if completed.returncode != 0:
+            raise RuntimeError(completed.stderr.strip() or completed.stdout.strip() or "failed to convert AI to PDF")
+        pdf_path = converted_dir / f"{source.stem}.pdf"
+
+    image_dir = workdir / "rasterized"
+    image_dir.mkdir(parents=True, exist_ok=True)
+    image_path = image_dir / f"{source.stem}.png"
+
+    if shutil.which("pdftoppm"):
+        command = ["pdftoppm", "-png", "-singlefile", "-r", "220", str(pdf_path), str(image_path.with_suffix(""))]
+        completed = subprocess.run(command, capture_output=True, text=True, check=False)
+        if completed.returncode != 0:
+            raise RuntimeError(completed.stderr.strip() or completed.stdout.strip() or "failed to rasterize PDF")
+        return image_path
+
+    if shutil.which("magick"):
+        command = ["magick", "-density", "220", str(pdf_path), "-quality", "100", str(image_path)]
+        completed = subprocess.run(command, capture_output=True, text=True, check=False)
+        if completed.returncode != 0:
+            raise RuntimeError(completed.stderr.strip() or completed.stdout.strip() or "failed to rasterize PDF")
+        return image_path
+
+    raise RuntimeError("Neither pdftoppm nor magick is available for PDF rasterization")
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description="Detect text-line and outer rectangle boxes from a label image.")
+    parser.add_argument("input", type=Path, help="Path to an image, PDF, or AI file")
+    parser.add_argument("--output-dir", type=Path, default=Path("outputs/layout_boxes"), help="Directory for preview and JSON")
+    args = parser.parse_args()
+
+    source = args.input.expanduser().resolve()
+    output_dir = args.output_dir.expanduser().resolve()
+    output_dir.mkdir(parents=True, exist_ok=True)
+
+    raster_path = ensure_raster_image(source, output_dir)
+    annotated, boxes = process_image(raster_path)
+
+    preview_path = output_dir / f"{source.stem}.boxed.png"
+    json_path = output_dir / f"{source.stem}.boxes.json"
+    cv2.imwrite(str(preview_path), annotated)
+    json_path.write_text(
+        json.dumps(
+            {
+                "source": str(source),
+                "raster_path": str(raster_path),
+                "preview_path": str(preview_path),
+                "boxes": [box.to_dict() for box in boxes],
+            },
+            ensure_ascii=False,
+            indent=2,
+        ),
+        encoding="utf-8",
+    )
+
+    print(json.dumps({"preview_path": str(preview_path), "json_path": str(json_path)}, ensure_ascii=False, indent=2))
+
+
+if __name__ == "__main__":
+    main()

scripts/detect_regions.py

@@ -0,0 +1,109 @@
+#!/usr/bin/env python3
+"""Standalone CLI: detect main regions in a label image via Qwen VL, then crop.
+
+Usage
+-----
+    python scripts/detect_regions.py <image_path> [--model MODEL] [--out OUT_DIR] [--key KEY]
+
+Example
+-------
+    python scripts/detect_regions.py data/sample.png
+    python scripts/detect_regions.py data/sample.png --model qwen2.5-vl-72b-instruct
+"""
+from __future__ import annotations
+
+import argparse
+import json
+import logging
+import sys
+from pathlib import Path
+
+# ── make sure the project root is on sys.path ──────────────────────────────
+ROOT = Path(__file__).resolve().parent.parent
+sys.path.insert(0, str(ROOT))
+
+logging.basicConfig(
+    level=logging.INFO,
+    format="%(asctime)s [%(levelname)s] %(name)s – %(message)s",
+)
+logger = logging.getLogger("detect_regions")
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(
+        description="Detect & crop main regions in a label image using Qwen VL"
+    )
+    parser.add_argument("image", help="Path to input image (PNG/JPEG)")
+    parser.add_argument(
+        "--model",
+        default="qwen2.5-vl-7b-instruct",
+        help="DashScope model ID (default: qwen2.5-vl-7b-instruct)",
+    )
+    parser.add_argument(
+        "--out",
+        default=None,
+        help="Output directory for cropped regions (default: <image_dir>/regions/)",
+    )
+    parser.add_argument("--key", default=None, help="DASHSCOPE_API_KEY (overrides env)")
+    parser.add_argument(
+        "--api-max-side", type=int, default=1024,
+        help="Max side length (px) of image sent to API (default: 1024). "
+             "Crop is always done on the original full-res file.",
+    )
+    parser.add_argument("--no-crop", action="store_true", help="Only print coords, don't crop")
+    parser.add_argument("--split", action="store_true",
+                        help="Save each detected region separately (default: merge into one)")
+    args = parser.parse_args()
+
+    image_path = Path(args.image).expanduser().resolve()
+    if not image_path.exists():
+        parser.error(f"Image not found: {image_path}")
+
+    output_dir = Path(args.out).expanduser().resolve() if args.out else image_path.parent / "regions"
+
+    from backend.app.region_detector import detect_regions, crop_and_save, merge_regions
+
+    logger.info("Image: %s", image_path)
+    logger.info("Model: %s", args.model)
+
+    regions, raw_response = detect_regions(
+        image_path,
+        api_key=args.key or None,
+        model=args.model,
+        api_max_side=args.api_max_side,
+    )
+
+    if not regions:
+        logger.error("No regions detected. Raw model response:\n%s", raw_response)
+        sys.exit(1)
+
+    print("\n── Detected regions ──────────────────────────────────────")
+    for i, r in enumerate(regions, 1):
+        print(f"  {i:02d}. [{r.label}]  bbox=({r.x1},{r.y1})-({r.x2},{r.y2})  "
+              f"size={r.width}×{r.height}px")
+    print()
+
+    coords_json = [
+        {"label": r.label, "bbox": [r.x1, r.y1, r.x2, r.y2]}
+        for r in regions
+    ]
+    print("JSON:")
+    print(json.dumps(coords_json, ensure_ascii=False, indent=2))
+
+    if not args.no_crop:
+        if args.split:
+            save_regions = regions
+        else:
+            merged = merge_regions(regions)
+            save_regions = [merged]
+            print(f"\n── Merged bbox: ({merged.x1},{merged.y1})-({merged.x2},{merged.y2})"
+                  f"  size={merged.width}×{merged.height}px")
+
+        results = crop_and_save(image_path, save_regions, output_dir)
+        print(f"\n── Cropped file(s) saved to: {output_dir} ──")
+        for item in results:
+            print(f"  • {Path(item['path']).name}  ← {item['label']}")
+
+
+if __name__ == "__main__":
+    main()

scripts/start_backend.sh

@@ -0,0 +1,7 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+
+cd "$ROOT_DIR"
+python3 -m uvicorn backend.app.main:app --host 127.0.0.1 --port 8010 --reload

scripts/start_dev.sh

@@ -0,0 +1,80 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+BACKEND_PORT="${BACKEND_PORT:-8010}"
+FRONTEND_PORT="${FRONTEND_PORT:-5173}"
+
+port_pid() {
+  lsof -tiTCP:"$1" -sTCP:LISTEN 2>/dev/null | head -n 1 || true
+}
+
+port_cmd() {
+  local pid
+  pid="$(port_pid "$1")"
+  if [[ -z "$pid" ]]; then
+    return 0
+  fi
+  ps -p "$pid" -o command= 2>/dev/null || true
+}
+
+print_port_conflict() {
+  local port="$1"
+  local pid
+  local cmd
+  pid="$(port_pid "$port")"
+  cmd="$(port_cmd "$port")"
+  echo "Port $port is already in use." >&2
+  if [[ -n "$pid" ]]; then
+    echo "  PID: $pid" >&2
+  fi
+  if [[ -n "$cmd" ]]; then
+    echo "  CMD: $cmd" >&2
+  fi
+}
+
+cleanup() {
+  if [[ -n "${BACKEND_PID:-}" ]]; then
+    kill "$BACKEND_PID" 2>/dev/null || true
+  fi
+  if [[ -n "${FRONTEND_PID:-}" ]]; then
+    kill "$FRONTEND_PID" 2>/dev/null || true
+  fi
+}
+
+trap cleanup EXIT INT TERM
+
+if [[ -n "$(port_pid "$BACKEND_PORT")" ]]; then
+  print_port_conflict "$BACKEND_PORT"
+  echo "Set BACKEND_PORT to another port or stop the existing process first." >&2
+  exit 1
+fi
+
+if [[ -n "$(port_pid "$FRONTEND_PORT")" ]]; then
+  print_port_conflict "$FRONTEND_PORT"
+  echo "Set FRONTEND_PORT to another port or stop the existing process first." >&2
+  exit 1
+fi
+
+cd "$ROOT_DIR"
+# Prefer miniconda Python (has all project deps); fall back to conda/system python3
+PYTHON="${CONDA_PREFIX:-}/bin/python"
+if [[ ! -x "$PYTHON" ]]; then
+  PYTHON="$(command -v /Users/icemilk/miniconda3/bin/python 2>/dev/null || command -v python3)"
+fi
+"$PYTHON" -m uvicorn backend.app.main:app --host 127.0.0.1 --port "$BACKEND_PORT" --reload \
+  > >(sed 's/^/[backend] /') \
+  2> >(sed 's/^/[backend] /' >&2) &
+BACKEND_PID=$!
+
+cd "$ROOT_DIR/frontend"
+npm run dev -- --host 127.0.0.1 --port "$FRONTEND_PORT" \
+  > >(sed 's/^/[frontend] /') \
+  2> >(sed 's/^/[frontend] /' >&2) &
+FRONTEND_PID=$!
+
+echo "Backend:  http://127.0.0.1:${BACKEND_PORT}"
+echo "Frontend: http://127.0.0.1:${FRONTEND_PORT}"
+echo "Press Ctrl+C to stop both services."
+
+wait "$BACKEND_PID" "$FRONTEND_PID"

scripts/start_frontend.sh

@@ -0,0 +1,7 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+
+cd "$ROOT_DIR/frontend"
+npm run dev -- --host 127.0.0.1 --port 5173