lazylabel-gui 1.1.6

An image segmentation GUI for generating ML ready mask tensors and annotations.

LazyLabel is an intuitive, AI-assisted image segmentation tool built with a modern, modular architecture. It uses Meta's Segment Anything Model (SAM) for quick, precise mask generation, alongside advanced polygon editing for fine-tuned control. Features comprehensive model management, customizable hotkeys, and outputs in clean, one-hot encoded .npz format for easy machine learning integration.

Inspired by LabelMe and Segment-Anything-UI.

---

✨ Core Features

AI-Powered Segmentation

• Generate masks with simple left-click (positive) and right-click (negative) interactions
• Multiple SAM model support with easy switching
• Custom model loading from any directory

Advanced Editing Tools

• Vector Polygon Tool: Full control to draw, edit, and reshape polygons
• Vertex Editing: Drag vertices or move entire shapes with precision
• Selection & Merging: Select, merge, and re-order segments intuitively

Professional Workflow

• Customizable Hotkeys: Personalize keyboard shortcuts for all functions
• Advanced Class Management: Assign multiple segments to single class IDs
• Smart I/O: Load existing .npz masks; save as clean, one-hot encoded outputs
• Interactive UI: Color-coded segments, sortable lists, and hover highlighting

Modern Architecture

• Modular Design: Clean, maintainable codebase with separated concerns
• Model Management: Dedicated model storage and switching system
• Persistent Settings: User preferences saved between sessions

---

🚀 Getting Started

Prerequisites

Python 3.10+

Installation

For Users via PyPI

1. Install LazyLabel directly:

   pip install lazylabel-gui
   

2. Run the application:

   lazylabel-gui
   

For Developers (from Source)

1. Clone the repository:

   git clone https://github.com/dnzckn/LazyLabel.git
   cd LazyLabel
   

2. Install in editable mode:

   pip install -e .
   

3. Run the application:

   lazylabel-gui
   

Model Management

• Default Storage: Models are stored in src/lazylabel/models/ directory
• Custom Models: Click "Browse Models" to select custom model folders
• Model Switching: Use the dropdown to switch between available models
• Auto-Detection: Application automatically detects all .pth files in selected directories

Note: On the first run, the application will automatically download the SAM model checkpoint (~2.5 GB) from Meta's repository to the models directory. This is a one-time download.

---

⌨️ Controls & Keybinds

💡 Tip: All hotkeys are fully customizable! Click the "Hotkeys" button in the control panel to personalize your shortcuts.

Modes

Key	Action
1	Enter Point Mode (for AI segmentation).
2	Enter Polygon Drawing Mode.
3	Enter Bounding Box Mode.
E	Toggle Selection Mode to select existing segments.
R	Enter Edit Mode for selected polygons (drag shape or vertices).
Q	Toggle Pan Mode (click and drag the image).

Actions

Key(s)	Action
L-Click	Add positive point (Point Mode) or polygon vertex.
R-Click	Add negative point (Point Mode).
Ctrl + Z	Undo last action.
Ctrl + Y / Ctrl + Shift + Z	Redo last action.
Spacebar	Finalize and save current AI segment.
Enter	Save final mask for the current image to a .npz file.
M	Merge selected segments into a single class.
V / Delete / Backspace	Delete selected segments.
C	Clear temporary points/vertices.
W/A/S/D	Pan image.
Scroll Wheel	Zoom-in or -out.

---

📦 Output Format

LazyLabel saves your work as a compressed NumPy array (.npz) with the same name as your image file.

The file contains a single data key, 'mask', holding a one-hot encoded tensor with the shape (H, W, C):

• H: Image height.
• W: Image width.
• C: Total unique classes.

Each channel is a binary mask for a class, combining all assigned segments into a clean, ML-ready output.

---

🏗️ Architecture

LazyLabel features a modern, modular architecture designed for maintainability and extensibility:

• Modular Design: Clean separation between UI, business logic, and configuration
• Signal-Based Communication: Loose coupling between components using PyQt signals
• Persistent Configuration: User settings and preferences saved between sessions
• Extensible Model System: Easy integration of new SAM models and types

For detailed technical documentation, see ARCHITECTURE.md.

---

⌨️ Hotkey Customization

LazyLabel includes a comprehensive hotkey management system:

• Full Customization: Personalize keyboard shortcuts for all 27+ functions
• Category Organization: Hotkeys organized by function (Modes, Actions, Navigation, etc.)
• Primary & Secondary Keys: Set multiple shortcuts for the same action
• Persistent Settings: Custom hotkeys saved between sessions
• Conflict Prevention: System prevents duplicate key assignments

For complete hotkey documentation, see HOTKEY_FEATURE.md.

Development

Code Quality

This project uses Ruff for linting and formatting:

# Activate virtual environment first
& e:\venv\lazylabel\Scripts\Activate.ps1

# Run linter
ruff check .

# Fix auto-fixable issues
ruff check --fix .

# Format code
ruff format .

# Check if code is properly formatted
ruff format --check .

Testing

Run tests using pytest:

# Run all tests
python -m pytest

# Run tests with coverage
python -m pytest --cov=lazylabel --cov-report=html --cov-report=term-missing

# Run specific test file
python -m pytest tests/unit/ui/test_undo_redo.py -v

The HTML coverage report will be generated in htmlcov/ directory.

---

☕ Support LazyLabel

If you found LazyLabel helpful, consider supporting the project!