Initial commit: Invoice field extraction system using YOLO + OCR

Features:
- Auto-labeling pipeline: CSV values -> PDF search -> YOLO annotations
- Flexible date matching: year-month match, nearby date tolerance
- PDF text extraction with PyMuPDF
- OCR support for scanned documents (PaddleOCR)
- YOLO training and inference pipeline
- 7 field types: InvoiceNumber, InvoiceDate, InvoiceDueDate, OCR, Bankgiro, Plusgiro, Amount

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

scripts/run_autolabel.sh

@@ -0,0 +1,55 @@
+#!/bin/bash
+# 自动标注运行脚本
+# 使用方法: bash scripts/run_autolabel.sh
+
+set -e
+
+# 项目根目录
+PROJECT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+cd "$PROJECT_DIR"
+
+# 激活虚拟环境
+if [ -f "venv/bin/activate" ]; then
+    source venv/bin/activate
+else
+    echo "错误: 虚拟环境不存在，请先运行 setup_wsl.sh"
+    exit 1
+fi
+
+# 默认参数
+CSV_FILE="${CSV_FILE:-data/structured_data/invoices.csv}"
+PDF_DIR="${PDF_DIR:-data/raw_pdfs}"
+OUTPUT_DIR="${OUTPUT_DIR:-data/dataset}"
+REPORT_FILE="${REPORT_FILE:-reports/autolabel_report.jsonl}"
+DPI="${DPI:-300}"
+MIN_CONFIDENCE="${MIN_CONFIDENCE:-0.7}"
+
+# 显示配置
+echo "=========================================="
+echo "自动标注配置"
+echo "=========================================="
+echo "CSV 文件:    $CSV_FILE"
+echo "PDF 目录:    $PDF_DIR"
+echo "输出目录:    $OUTPUT_DIR"
+echo "报告文件:    $REPORT_FILE"
+echo "DPI:         $DPI"
+echo "最小置信度:  $MIN_CONFIDENCE"
+echo "=========================================="
+echo ""
+
+# 创建必要目录
+mkdir -p "$(dirname "$REPORT_FILE")"
+mkdir -p "$OUTPUT_DIR"
+
+# 运行自动标注
+python -m src.cli.autolabel \
+    --csv "$CSV_FILE" \
+    --pdf-dir "$PDF_DIR" \
+    --output "$OUTPUT_DIR" \
+    --report "$REPORT_FILE" \
+    --dpi "$DPI" \
+    --min-confidence "$MIN_CONFIDENCE" \
+    --verbose
+
+echo ""
+echo "完成! 数据集已生成到: $OUTPUT_DIR"

scripts/run_train.sh

@@ -0,0 +1,67 @@
+#!/bin/bash
+# 训练运行脚本
+# 使用方法: bash scripts/run_train.sh
+
+set -e
+
+# 项目根目录
+PROJECT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+cd "$PROJECT_DIR"
+
+# 激活虚拟环境
+if [ -f "venv/bin/activate" ]; then
+    source venv/bin/activate
+else
+    echo "错误: 虚拟环境不存在，请先运行 setup_wsl.sh"
+    exit 1
+fi
+
+# 默认参数
+DATA_YAML="${DATA_YAML:-data/dataset/dataset.yaml}"
+MODEL="${MODEL:-yolov8s.pt}"
+EPOCHS="${EPOCHS:-100}"
+BATCH_SIZE="${BATCH_SIZE:-16}"
+IMG_SIZE="${IMG_SIZE:-1280}"
+DEVICE="${DEVICE:-0}"
+
+# 检查数据集是否存在
+if [ ! -f "$DATA_YAML" ]; then
+    echo "错误: 数据集配置文件不存在: $DATA_YAML"
+    echo "请先运行自动标注: bash scripts/run_autolabel.sh"
+    exit 1
+fi
+
+# 显示配置
+echo "=========================================="
+echo "训练配置"
+echo "=========================================="
+echo "数据集:      $DATA_YAML"
+echo "基础模型:    $MODEL"
+echo "Epochs:      $EPOCHS"
+echo "Batch Size:  $BATCH_SIZE"
+echo "图像尺寸:    $IMG_SIZE"
+echo "设备:        $DEVICE"
+echo "=========================================="
+echo ""
+
+# 检查 GPU
+if command -v nvidia-smi &> /dev/null; then
+    echo "GPU 状态:"
+    nvidia-smi --query-gpu=name,memory.used,memory.total --format=csv,noheader
+    echo ""
+else
+    echo "警告: 未检测到 GPU，将使用 CPU 训练 (较慢)"
+    DEVICE="cpu"
+fi
+
+# 运行训练
+python -m src.cli.train \
+    --data "$DATA_YAML" \
+    --model "$MODEL" \
+    --epochs "$EPOCHS" \
+    --batch "$BATCH_SIZE" \
+    --imgsz "$IMG_SIZE" \
+    --device "$DEVICE"
+
+echo ""
+echo "训练完成! 模型保存在: runs/train/invoice_fields/weights/"

scripts/setup_wsl.sh

@@ -0,0 +1,80 @@
+#!/bin/bash
+# WSL 环境安装脚本
+# 使用方法: bash scripts/setup_wsl.sh
+
+set -e
+
+echo "=========================================="
+echo "Invoice Master POC v2 - WSL 安装脚本"
+echo "=========================================="
+
+# 检查是否在 WSL 中运行
+if ! grep -qi microsoft /proc/version 2>/dev/null; then
+    echo "警告: 未检测到 WSL 环境，请在 WSL 中运行此脚本"
+    echo "提示: 在 Windows 终端中输入 'wsl' 进入 WSL"
+    exit 1
+fi
+
+echo ""
+echo "[1/5] 更新系统包..."
+sudo apt update
+
+echo ""
+echo "[2/5] 安装系统依赖..."
+sudo apt install -y \
+    python3.10 \
+    python3.10-venv \
+    python3-pip \
+    libgl1-mesa-glx \
+    libglib2.0-0 \
+    libsm6 \
+    libxrender1 \
+    libxext6 \
+    libgomp1
+
+echo ""
+echo "[3/5] 创建 Python 虚拟环境..."
+if [ -d "venv" ]; then
+    echo "虚拟环境已存在，跳过创建"
+else
+    python3 -m venv venv
+fi
+
+echo ""
+echo "[4/5] 激活虚拟环境并安装依赖..."
+source venv/bin/activate
+pip install --upgrade pip
+
+echo ""
+echo "安装 Python 依赖包..."
+pip install -r requirements.txt
+
+echo ""
+echo "[5/5] 验证安装..."
+python3 -c "import fitz; print(f'PyMuPDF: {fitz.version}')"
+python3 -c "from ultralytics import YOLO; print('Ultralytics: OK')"
+python3 -c "from paddleocr import PaddleOCR; print('PaddleOCR: OK')"
+
+echo ""
+echo "=========================================="
+echo "安装完成!"
+echo "=========================================="
+echo ""
+echo "使用方法:"
+echo "  1. 激活虚拟环境: source venv/bin/activate"
+echo "  2. 运行自动标注: python -m src.cli.autolabel --help"
+echo "  3. 训练模型:     python -m src.cli.train --help"
+echo "  4. 推理:         python -m src.cli.infer --help"
+echo ""
+
+# 检查 GPU
+echo "检查 GPU 支持..."
+if command -v nvidia-smi &> /dev/null; then
+    echo "检测到 NVIDIA GPU:"
+    nvidia-smi --query-gpu=name,memory.total --format=csv,noheader
+    echo ""
+    echo "提示: 运行以下命令启用 GPU 加速:"
+    echo "  pip install paddlepaddle-gpu"
+else
+    echo "未检测到 GPU，将使用 CPU 模式"
+fi