LabelCraft 2.0.4

LabelCraft - A modern graphical image annotation tool based on labelImg

LabelCraft - Intelligent Image Annotation Tool

Version 2.0.0 - A modern image annotation tool with project management, developed based on labelImg

中文文档 | English

LabelCraft is a professional graphical image annotation tool with advanced project management capabilities. It's designed for efficient data preparation in deep learning tasks such as object detection and image classification.

This project is a major evolution of the open-source project labelImg, adding project-based workflow, multi-format support, and intelligent annotation features. Special thanks to the original author TzuTa Lin for his pioneering work.

✨ Key Features

🎯 Core Annotation

• Rectangle bounding box annotation
• Polygon annotation (coming soon)
• Real-time preview and editing
• Undo/Redo support

📁 Project Management (v2.0 New!)

• Create and manage annotation projects
• Project configuration persistence (.labelcraft files)
• Recent projects quick access
• Automatic annotation directory management
• Project metadata tracking

🔄 Multi-Format Support

Support 5 major annotation formats with seamless conversion:

• PASCAL VOC (XML) - Traditional format for Faster R-CNN, SSD, etc.
• YOLO (TXT) - For YOLO series models (v5, v8, v10)
• CreateML (JSON) - For Apple CreateML framework
• COCO (JSON) - Microsoft COCO dataset standard
• CSV (CSV) - Universal format for data analysis

🌐 Advanced Internationalization

Dynamic language switching without restart:

• English
• 简体中文 (Simplified Chinese)
• 繁體中文 (Traditional Chinese)
• 日本語 (Japanese)
• Deutsch (German)
• Français (French)

⚡ Smart Workflow

• Pending annotation queue management
• Completed annotations list
• Auto-save mode
• Default label for batch annotation
• Verification mode for quality control
• Copy previous frame annotations

🛠️ Developer Tools

• Built-in format converter (all 5 formats)
• Command-line interface
• Python API for integration
• Customizable shortcuts
• Brightness adjustment

📸 Screenshot

🚀 Quick Start

Installation via pip (Recommended)

pip install labelcraft

Launch from command line:

labelcraft

Or specify an image directory:

labelcraft /path/to/images

One-Click Launch (From Source)

Linux/macOS:

chmod +x start.sh
./start.sh

Windows:

start.bat

The script will automatically:

1. Check Python environment
2. Create virtual environment
3. Install dependencies
4. Compile resources
5. Launch application

Manual Installation

# Clone repository
git clone https://github.com/syd168/LabelCraft.git
cd LabelCraft

# Create virtual environment
python3 -m venv venv
source venv/bin/activate  # Linux/macOS
# or
venv\Scripts\activate     # Windows

# Install dependencies
pip install -r requirements.txt

# Compile resources (Linux/macOS)
make pyside6
# or Windows
pyside6-rcc -o libs/resources.py resources.qrc

# Run
python main.py

📖 Usage Guide

Standard Workflow (Project-Based)

Step 1: Create a New Project

1. Menu: File → New Project or press Ctrl+N
2. Fill in project information:
   • Project Name: Your project name
   • Project Location: Directory to store project files
   • Output Format: Choose annotation format (VOC/YOLO/CreateML/COCO/CSV)
   • Labels: Add your object categories
3. Click "Create Project"

The system will automatically create:

• Project file: {project_name}.labelcraft
• Annotations directory: {project_dir}/annotations/

Step 2: Add Images to Project

1. Menu: File → Open Dir or press Ctrl+U
2. Select the directory containing your images
3. Images will be loaded into the pending queue

You can also:

• Drag and drop images directly
• Use File → Add Images to add specific files

Step 3: Annotate Images

1. Click "Create RectBox" button or press W
2. Draw bounding box on the object
3. Enter label name (or use default label)
4. Press Enter to confirm
5. Save with Ctrl+S or enable auto-save

Tips:

• Use arrow keys to fine-tune box position
• Hold Ctrl while drawing for perfect squares
• Right-click box to edit properties
• Use verification mode (Space) to mark completed images

Step 4: Manage Annotations

View Status:

• Left panel: Pending images queue
• Right panel: Current image annotations
• Window title shows progress: (5/100)

Navigation:

• Next image: D or →
• Previous image: A or ←
• Jump to specific image from file list

Step 5: Export & Convert

Export Annotations:

• Menu: File → Export Annotations
• Choose export location and format

Convert Formats: Use the built-in converter programmatically:

from libs.annotation_converter import AnnotationConverter

converter = AnnotationConverter()

# Convert entire project from VOC to YOLO
converter.convert(
    input_dir='path/to/voc_annotations',
    input_format='voc',
    output_format='yolo',
    output_dir='path/to/yolo_output'
)

Or via command line:

python -m libs.annotation_converter \
    --input /path/to/input \
    --input_format voc \
    --output_format yolo \
    --output /path/to/output

Legacy Mode (Quick Annotation)

For simple tasks without project management:

# Open single image
python main.py image.jpg

# Open directory directly
python main.py /path/to/images

# With predefined labels
python main.py /path/to/images data/predefined_classes.txt

⌨️ Keyboard Shortcuts

File Operations

Shortcut	Action
Ctrl+N	New Project
Ctrl+O	Open Project
Ctrl+U	Open Directory
Ctrl+S	Save Annotation
Ctrl+E	Edit Project
Ctrl+Shift+C	Close Project

Annotation

Shortcut	Action
W	Create RectBox
Ctrl+J	Edit Mode
Delete	Delete Selected Box
Ctrl+D	Duplicate Box
Ctrl+V	Copy Previous Frame
Space	Verify Image

Navigation

Shortcut	Action
D / →	Next Image
A / ←	Previous Image
Ctrl++	Zoom In
Ctrl+-	Zoom Out
Ctrl+F	Fit Window
Ctrl+Shift+F	Fit Width

View

Shortcut	Action
Ctrl+T	Toggle Toolbar
Ctrl+H	Hide All Boxes
Ctrl+A	Show All Boxes
Ctrl+Shift++	Increase Brightness
Ctrl+Shift+-	Decrease Brightness
Ctrl+Shift+=	Reset Brightness

🔧 Advanced Configuration

Predefined Labels

Create a text file with one label per line:

person
car
dog
cat
bicycle

Load at startup:

python main.py data/predefined_classes.txt

Or at runtime: File → Load Predefined Classes (Ctrl+Shift+L)

Custom Settings

Settings are automatically saved to system config:

• Recently used directories
• Window size and position
• Default annotation format
• Brush colors
• Language preference

Command Line Arguments

python main.py [IMAGE_PATH] [PREDEFINED_CLASSES] [SAVE_DIR]

Examples:

# Open specific image
python main.py images/test.jpg

# With predefined classes
python main.py images/ data/classes.txt

# Specify save directory
python main.py images/ classes.txt annotations/

🏗️ Project Structure

LabelCraft/
├── main.py                 # Application entry point
├── labelcraft_ui.py        # Main UI implementation
├── libs/                   # Core libraries
│   ├── project.py          # Project management
│   ├── canvas.py           # Drawing canvas
│   ├── shape.py            # Shape classes
│   ├── annotation_converter.py  # Format converter
│   ├── i18n_engine.py      # Internationalization
│   ├── pascal_voc_io.py    # VOC format I/O
│   ├── yolo_io.py          # YOLO format I/O
│   ├── create_ml_io.py     # CreateML I/O
│   ├── coco_io.py          # COCO I/O
│   └── csv_io.py           # CSV I/O
├── locales/                # Translation files
│   ├── en.json
│   ├── zh-CN.json
│   ├── zh-TW.json
│   ├── ja-JP.json
│   ├── de-DE.json
│   └── fr-FR.json
├── resources/              # Icons and assets
├── doc/                    # Documentation
│   ├── tutorial.md
│   ├── tutorial_zh-CN.md
│   └── TRANSLATION_GUIDE.md
└── build-tools/            # Build scripts

🌍 Adding New Language

1. Create new file in locales/ (e.g., es-ES.json)
2. Copy existing JSON as template
3. Translate all values (keep keys unchanged)
4. Restart application - new language auto-detected

See TRANSLATION_GUIDE.md for details.

❓ FAQ

Q: What's the difference between v1.x and v2.0?

A: Version 2.0 introduces:

• ✅ Project-based workflow (vs. file-based in v1.x)
• ✅ Support for COCO and CSV formats
• ✅ Built-in format converter
• ✅ Dynamic language switching
• ✅ Enhanced UI with better organization
• ✅ Improved annotation management

Q: Can I still use it without creating a project?

A: Yes! You can use legacy mode by opening directories directly. However, projects provide better organization and persistence.

Q: How do I convert my existing annotations?

A: Use the built-in converter:

from libs.annotation_converter import AnnotationConverter

converter = AnnotationConverter()
converter.convert('old_annotations/', 'voc', 'yolo', 'new_annotations/')

Q: Where are my annotations saved?

A:

• With Project: In {project_dir}/annotations/
• Without Project: Same directory as images
• Filename matches image with appropriate extension (.xml, .txt, .json, .csv)

Q: Can I change the output format mid-project?

A: Yes! Edit project (Ctrl+E) and change the format. Existing annotations will be migrated if needed.

Q: How do I backup my project?

A: Simply copy the entire project directory including:

• .labelcraft project file
• annotations/ directory
• Your images (if stored in project)

Q: Is there an API for programmatic use?

A: Yes! You can integrate LabelCraft into your Python code:

from libs.project import Project
from libs.annotation_converter import AnnotationConverter

# Create project
project = Project(name="MyProject", project_dir="/path/to/project")
project.save("/path/to/project/myproject.labelcraft")

# Convert formats
converter = AnnotationConverter()
converter.convert("input/", "voc", "yolo", "output/")

🤝 Contributing

Contributions are welcome! Please follow these steps:

1. Fork the repository
2. Create feature branch: git checkout -b feature/amazing-feature
3. Commit changes: git commit -m 'Add amazing feature'
4. Push to branch: git push origin feature/amazing-feature
5. Open a Pull Request

Development Setup

# Clone and setup
git clone https://github.com/syd168/LabelCraft.git
cd LabelCraft
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

# Run tests
python -m pytest tests/

# Compile resources after UI changes
pyside6-rcc -o libs/resources.py resources.qrc

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• labelImg - Original project by TzuTa Lin
• All contributors who improved LabelCraft
• The open-source community for inspiration and support

📮 Contact & Support

• Project Homepage: https://github.com/syd168/LabelCraft
• Issue Tracker: https://github.com/syd168/LabelCraft/issues
• Documentation: See doc/ directory
• Email: [e-mail]

---

Happy Annotating! 🎉

Made with ❤️ for the computer vision community