AlchemyAnnotate is a free, open-source, desktop image annotation tool built with Python and PySide6. It provides a fast and intuitive interface for creating bounding box annotations used in training object detection models such as YOLO, Faster R-CNN, and SSD.
Export your annotations in YOLO, Pascal VOC, and COCO formats with a single click.
- Bounding Box Annotation - Draw, select, edit, and delete bounding boxes directly on images
- Multi-Format Export - Save annotations in YOLO, Pascal VOC (XML), and COCO (JSON) formats
- Separate Annotation Folders - Each format is stored in its own folder (
annotations_yolo/,annotations_voc/,annotations_coco/) to prevent overwrites - Class Management - Create, delete, and assign annotation classes through the UI with automatic color coding
- Autosave - Annotations are saved automatically after every change
- Format Switching - Switch between YOLO, VOC, and COCO at any time with automatic conversion of existing annotations
- Existing Annotation Detection - Automatically detects and loads previously saved annotations when opening a folder
- Project Session File - Remembers your last opened image, selected format, class list, and preferences across sessions
- Image Navigation - Browse images using sidebar, keyboard shortcuts, or arrow keys
- Zoom and Pan - Zoom with Ctrl+Scroll and pan with middle-click for precise annotation on large images
- Sidebar Status - Visual indicators show which images are labeled and which are still pending
| Format | Output Folder | File Type | Used By |
|---|---|---|---|
| YOLO | annotations_yolo/ |
.txt per image + classes.txt |
YOLOv5, YOLOv8, Ultralytics |
| Pascal VOC | annotations_voc/ |
.xml per image |
Faster R-CNN, SSD, Detectron |
| COCO | annotations_coco/ |
Single annotations.json |
COCO API, Detectron2, MMDetection |
JPG, JPEG, PNG, BMP, TIFF, WEBP
pip install alchemyannotategit clone https://github.com/kouya-marino/AlchemyAnnotate.git
cd AlchemyAnnotate
pip install .git clone https://github.com/kouya-marino/AlchemyAnnotate.git
cd AlchemyAnnotate
pip install -e ".[dev]"- Python 3.10 or higher
- PySide6 (installed automatically)
python run.pyOr run as a Python module:
python -m alchemyannotateOr if installed as a package:
alchemyannotate- Open a Folder - Click
File > Open Folderor pressCtrl+Oto select a folder containing your images - Create Classes - In the right panel, type a class name (e.g., "car", "person") and click Add
- Select a Class - Click on a class in the class list to make it active
- Draw Bounding Boxes - Left-click and drag on the image to draw a box around an object
- Navigate Images - Use the sidebar, or press
A/Dto go to the previous / next image - Edit Boxes - Click a box to select it, then change its class from the dropdown or press
Deleteto remove it - Save - Annotations autosave after every change. Press
Ctrl+Sto force a manual save - Switch Format - Use the format selector in the status bar to switch between YOLO, VOC, and COCO
| Shortcut | Action |
|---|---|
Ctrl+O |
Open image folder |
Ctrl+S |
Save annotations |
Ctrl+Q |
Quit application |
Delete |
Delete selected bounding box |
A or Left Arrow |
Previous image |
D or Right Arrow |
Next image |
Ctrl+0 |
Fit image to window |
Ctrl+Scroll |
Zoom in / out |
Middle-click drag |
Pan the image |
When you open a folder, AlchemyAnnotate creates an alchemyannotate_project.json file inside it. This file stores:
- Selected annotation format
- Class list
- Last opened image
- Autosave preferences
- Recently used class
This allows you to resume your annotation session exactly where you left off.
AlchemyAnnotate/
├── alchemyannotate/
│ ├── models/ # Data models (BoundingBox, ImageAnnotation, ProjectConfig)
│ ├── views/ # UI components (Canvas, Sidebar, Panels, Dialogs)
│ ├── controllers/ # Application logic (App, Canvas, Navigation controllers)
│ ├── services/ # IO, autosave, format conversion, image loading
│ └── utils/ # Constants, geometry helpers
├── tests/ # Unit tests for models, IO formats, and store
├── run.py # Quick launcher
├── pyproject.toml # Build configuration
└── requirements.txt # Dependencies
pip install -e ".[dev]"
pytest tests/ -v| Package | Version | Purpose |
|---|---|---|
| PySide6 | >= 6.6 | Qt6 GUI framework |
No other external dependencies. XML and JSON handling use Python standard library modules.
- Use YOLO if you are training with Ultralytics YOLOv5, YOLOv8, or similar YOLO-based models
- Use Pascal VOC if you are training with Faster R-CNN, SSD, or frameworks that expect XML annotations
- Use COCO if you are using Detectron2, MMDetection, or the COCO evaluation API
Yes. AlchemyAnnotate will prompt you to export existing annotations to the new format when you switch.
Inside your image folder, in a subfolder named after the format:
annotations_yolo/for YOLO formatannotations_voc/for Pascal VOC formatannotations_coco/for COCO format
Yes. AlchemyAnnotate supports both bounding box and polygon annotations. Polygons export as COCO segmentation fields and YOLOv8-seg format. Note: Pascal VOC does not support polygons — a warning is shown when switching to VOC if polygon data exists.
This project is open source. See the LICENSE file for details.
Contributions are welcome. Please open an issue or submit a pull request on GitHub.
