git clone https://github.com/HR-DONK/circuit-recognition.git
cd circuit-recognition

# 安装依赖
pip install -r requirements.txt

{
  "mcpServers": {
    "circuit-recognition": {
      "command": "python",
      "args": ["src/mcp_server.py"],
      "cwd": "/path/to/circuit-recognition"
    }
  }
}

{
  "success": true,
  "components": [
    { "class": "resistor", "bbox": [120, 80, 60, 20], "confidence": 0.92 },
    { "class": "capacitor", "bbox": [120, 160, 40, 30], "confidence": 0.88 }
  ],
  "connections": [
    { "from": "R1.pin2", "to": "C1.pin1", "wire_segments": [...] }
  ],
  "nets": {
    "VCC": ["R1.pin1"],
    "GND": ["C1.pin2", "R2.pin1"],
    "NET_1": ["R1.pin2", "C1.pin1"]
  },
  "title_block": { "title": "RC 滤波电路" },
  "timing": {
    "yolo": 0.12, "ocr": 1.68, "total": 2.34
  }
}

输入图片 → YOLO元件检测 → SAM精细分割 → OCR文字识别 → 拓扑分析 → 融合输出
  0.1s        0.1s          按需触发       1.7s        0.2s        JSON

circuit-recognition/
├── src/
│   ├── mcp_server.py              # MCP 服务入口（4 个工具）
│   ├── pipeline/
│   │   ├── yolo_detector.py       # Stage 1: YOLO 元件检测
│   │   ├── sam_segmentor.py       # Stage 2: SAM 精细分割
│   │   ├── ocr_engine.py          # Stage 3: OCR 引擎（调度层）
│   │   ├── smart_filter.py        #     文字框多级过滤（size/conf/NMS/ROI/margin）
│   │   ├── backends/
│   │   │   ├── base.py            #     OCR 后端抽象基类
│   │   │   ├── rapidocr_backend.py #    RapidOCR 后端（默认，CPU 可用）
│   │   │   └── trt_backend.py     #     TensorRT 后端（GPU 加速）
│   │   ├── topology.py            # Stage 4: 拓扑分析
│   │   ├── fusion.py              # Stage 5: 融合 + 网表生成
│   │   └── preprocessor.py        # 图像预处理 + PDF 转换
│   └── cache.py                   # 多级缓存
├── config/
│   └── ocr_config.yaml            # OCR 参数配置
├── scripts/
│   └── build_trt_engines.py       # TensorRT 引擎构建（需 GPU）
├── tests/                         # 38 个测试
├── models/                        # 模型文件（自动下载）
└── requirements.txt

from src.pipeline.backends.base import OcrBackendBase
import numpy as np

class MyOcrBackend(OcrBackendBase):
    """接入你的 OCR 引擎"""

    def initialize(self) -> None:
        pass  # 初始化你的 OCR 模型

    def release(self) -> None:
        pass  # 释放资源

    @property
    def is_available(self) -> bool:
        return True

    def detect(self, image: np.ndarray) -> dict:
        # 文字区域检测
        # 返回 {"boxes": [...], "scores": [...], "cost_sec": 0.0, "img_hw": (h, w)}
        pass

    def recognize_batch(self, image: np.ndarray, boxes: list) -> list:
        # 批量文字识别
        # 返回 [{"text": "...", "confidence": 0.95, "bbox": [x,y,w,h]}, ...]
        pass

backend:
  default: myocr

Python 3.10+
├── torch + ultralytics      # YOLO 推理
├── segment-anything         # SAM 分割
├── rapidocr-onnxruntime     # OCR（默认后端）
├── opencv-python-headless   # 图像处理
├── PyMuPDF                  # PDF 支持
└── mcp                      # MCP 协议框架