UniLabel - Universal Object Detection Converter

通用目标检测标签格式转换工具

---

UniLabel is a powerful and user-friendly GUI tool designed to convert object detection dataset annotations between various formats. Built with PySide6, it employs an Intermediate Representation (IR) architecture to ensure accurate and robust conversions.

UniLabel 是一个基于 PySide6 开发的通用目标检测数据集格式转换工具。它采用中间表示（IR）架构，能够实现主流标注格式之间的相互转换，支持批量处理，操作简单高效。

---

🔄 Supported Matrix / 支持矩阵

Format	Import (Read)	Export (Write)	Notes / 备注
Pascal VOC (.xml)	✅	✅	Standard XML format / 标准 XML
YOLO (.txt)	✅	✅	Requires Images & classes.txt / 需图片和类别文件
MS COCO (.json)	✅	✅	Supports Batch Processing / 支持批量
LabelMe (.json)	✅	✅	Auto-detects images / 自动寻找图片

---

🛠️ Installation / 安装指南

1. Clone the repository / 克隆仓库

   git clone https://github.com/LCYvision/UniLabel.git
   cd UniLabel

2. Install dependencies / 安装依赖

   pip install -r requirements.txt

3. Run the application / 运行软件

   python main.py

---

📖 Usage / 使用说明

1. Import Data / 导入数据

• Select the source format from the dropdown menu / 从下拉菜单中选择源格式。
• Click "Select Folder/File" to load your dataset / 点击 "选择文件夹/文件" 来加载您的数据集

2. Export Settings / 导出设置

• Select the target format (VOC, YOLO, COCO, or LabelMe) / 选择目标格式（VOC、YOLO、COCO 或 LabelMe）
• Click "Select Output Path" to choose where to save the converted files / 点击 "选择保存路径" 来选择保存转换后文件的位置。

3. Start Conversion / 开始转换

• Click "Start Conversion" / 点击“开始转换”
• Wait for the progress bar to complete / 等待进度条完成

---

⚠️ NOTE / 注意

The author only conducted a small amount of testing. You can test using a few samples in the test_import directory of the project.

作者只进行了少量的测试，你可以使用项目中test_import目录中的少量样本进行测试。

---

📂 Project Structure / 项目结构

UniLabel/
├── main.py      	    # Entry point & GUI logic (程序入口 & 界面逻辑)
├── converters.py       # Importers & Exporters logic (核心转换算法)
├── ir_label.py         # Intermediate Representation Data Models (中间格式定义)
├── requirements.txt    # Python dependencies (依赖库)
├── test_import/        # Sample data for testing (测试样本)
└── README.md           # Documentation