Annotate Lab - Simplifying Image Annotation

Annotate Lab is an open-source application designed for image annotation, comprising two main components: the client and the server. The client, a React application, is responsible for the user interface where users perform annotations. On the other hand, the server, a Flask application, manages persisting the annotated changes and generating masked and annotated images, along with configuration settings. More information can be found in our documentation.

Table of Contents

• Project Structure
• Dependencies
• Setup and Installation
• Running the Application
  • Running the Client
  • Running the Server
  • Running using Docker
• Running Tests
  • Client Tests
  • Server Tests
• Code Formatting
• Usage
• Settings
• Configurations (Optional)
• Demo
• Auto Bounding Box Selection with Segment Anything Model (SAM)
• Outputs
• YOLO Format
  • Normalization process of YOLO annotations
• Troubleshooting
• Contributing
• License
• Reporting Security Issues
• Acknowledgment

Project Structure [documentation page]

```sh

annotation-lab/ ├── client/ │ ├── public/ │ ├── src/ │ ├── package.json │ ├── package-lock.json │ └── ... (other React app files) ├── server/ │ ├── db/ │ ├── tests/ │ ├── venv/ │ ├── app.py │ ├── requirements.txt │ └── ... (other Flask app files) ├── README.md ```

Client

• public/: Static files and the root HTML file.
• src/: React components and other frontend code.
• package.json: Contains client dependencies and scripts.

Server

• db/: Database-related files and handlers.
• venv/: Python virtual environment (not included in version control).
• tests/: Contains test files.
• app.py: Main Flask application file.
• requirements.txt: Contains server dependencies.

Dependencies [documentation page]

Client

• React
• Axios
• Other dependencies as listed in package.json

Server

• Flask
• Pandas
• Other dependencies as listed in requirements.txt

Setup and Installation [documentation page]

Client Setup

1. Navigate to the client directory: sh cd client
2. Install the dependencies: sh npm install ### Server Setup
3. Navigate to the server directory: sh cd server
4. Create and activate a virtual environment: ```sh python3 -m venv venv

source venv/bin/activate # On Windows use venv\Scripts\activate 3. Install the dependencies: sh pip install -r requirements.txt ```

Running the Application

Running the Client

1. Navigate to the client directory: sh cd client
2. Install the dependencies: sh npm start The application should now be running on http://localhost:5173.

Running the Server

1. Navigate to the server directory: sh cd server
2. Activate the virtual environment: sh source venv/bin/activate # On Windows use `venv\Scripts\activate`
3. Start the Flask application: sh flask run The server should now be running on http://localhost:5000.

Running using Docker

Navigate to the root directory and run the following command to start the application: sh docker-compose build docker-compose up -d #running in detached mode The application should be running on http://localhost.

Running Tests [documentation page]

Client Tests

The client tests are located in the client/src directory and utilize .test.js extensions. They are built using Jest and React Testing Library.

Install Dependencies:

bash cd client npm install

Run Tests:

bash npm test

This command launches the test runner in interactive watch mode. It runs all test files and provides feedback on test results.

Server Tests

The server tests are located in the server/tests directory and are implemented using unittest.

Install Dependencies:

bash cd ../server python -m venv venv source venv/bin/activate # On Windows use `venv\Scripts\activate` pip install -r requirements.txt

Run Tests:

```bash python3 -m unittest discover -s tests -p 'test_*.py'

```

This command discovers and runs all test files (test_*.py) in the server/tests directory using unittest.

Code Formatting [documentation page]

Client-side (Vite React Application)

• Code Formatter: Prettier
• Configuration File: .prettierrc
• Command: Run npm run format or yarn format to format client-side code using Prettier.

Server-side (Flask Application)

• Code Formatter: Black
• Configuration File: pyproject.toml
• Command: Run black . to format server-side code using Black.

Usage

1. Open your web browser and navigate to http://localhost:5173.
2. Use the user interface to upload and annotate images.
3. The annotations and other interactions will be handled by the Flask server running at http://localhost:5000.

Settings [documentation page]

One can configure the tools, tags, upload images and do many more from the settings.

Configurations (Optional) [documentation page]

You can customize various aspects of Annotate-Lab through configuration settings. To do this, modify the config.py file in the server directory or the config.js file in the client directory. ```python

config.py

MASKBACKGROUNDCOLOR = (0, 0, 0) # Black background for masks SAMMODELENABLED = False # Segment Anything Model for auto bounding box selection ```

Javascript // config.js const config = { SERVER_URL, // url of server UPLOAD_LIMIT: 500, // image upload limit OUTLINE_THICKNESS_CONFIG : { // outline thickness of tools POLYGON: 2, CIRCLE: 2, BOUNDING_BOX: 2 }, SAM_MODEL_ENABLED: false, // displays button that allows auto bounding box selection SHOW_CLASS_DISTRIBUTION: true // displays annotated class distribution bar chart };

Demo V2.0

Auto Bounding Box Selection with Segment Anything Model (SAM)[documentation page]

Selection of bounding box automatically is made possible with the Segment Anything Model (SAM). One can toggle this feature from the configuration of server and client. When enabled, a wand icon will appear in the toolbar. Clicking the wand icon will initiate auto-annotation and display the results

Outputs [documentation page]

Sample of annotated image along with its mask and settings is show below.

```json { "orange.png": { "configuration": [ { "image-name": "orange.png", "regions": [ { "region-id": "13371375927088525", "image-src": "http://127.0.0.1:5000/uploads/orange.png", "class": "Print", "comment": "", "tags": "", "points": [ [ 0.5863691595741748, 0.7210152721281337 ], [ 0.6782101128815677, 0.6587584627896123 ], [ 0.7155520389516067, 0.5731553499491453 ], [ 0.7286721751383771, 0.40065210740699225 ], [ 0.7518847237765094, 0.352662483541882 ], [ 0.6862840428426572, 0.2307428985872776 ], [ 0.6045355019866261, 0.1581099543590026 ], [ 0.533888614827093, 0.13476365085705708 ], [ 0.44204766151970004, 0.13476365085705708 ], [ 0.3441512607414899, 0.17886222413850975 ], [ 0.2957076809749529, 0.23852499975459276 ], [ 0.2523103074340969, 0.3163460114277445 ], [ 0.2129498988737856, 0.418810343464061 ], [ 0.20891293389324087, 0.5121955574718431 ], [ 0.22506079381541985, 0.6016897208959676 ], [ 0.2563472724146416, 0.6652435470957082 ], [ 0.30378161093604245, 0.7197182552669145 ], [ 0.3683730506247584, 0.7819750646054359 ], [ 0.4057149766947973, 0.8066183849686005 ], [ 0.46223248642242376, 0.776786997160559 ], [ 0.5308608910916844, 0.7586287611034903 ] ] } ], "color-map": { "Apple": [ 244, 67, 54 ], "Orange": [ 33, 150, 243 ] } } ] } }

```

YOLO Format [documentation page]

YOLO format is also supported by A.Lab. Below is an example of annotated ripe and unripe tomatoes. The entire dataset can be found on Kaggle. In this example, 0 represents ripe tomatoes and 1 represents unripe ones.

The label of the above image are as follows: 0 0.213673 0.474717 0.310212 0.498856 0 0.554777 0.540507 0.306350 0.433638 1 0.378432 0.681239 0.223970 0.268879

Applying the generated labels we get following results.

Normalization process of YOLO annotations [documentation page]

Example Conversion

To convert non-normalized bounding box coordinates (x_max, y_max, x_min, y_min) to YOLO format (x_center, y_center, width, height):

Image Credit: Leandro de Oliveira

```python

Assuming row contains your bounding box coordinates

row = {'xmax': 400, 'xmin': 200, 'ymax': 300, 'ymin': 100} class_id = 0 # Example class id (replace with actual class id)

Image dimensions

WIDTH = 640 # annotated image width HEIGHT = 640 # annotated image height

Calculate width and height of the bounding box

width = row['xmax'] - row['xmin'] height = row['ymax'] - row['ymin']

Calculate the center of the bounding box

xcenter = row['xmin'] + (width / 2) ycenter = row['ymin'] + (height / 2)

Normalize the coordinates

normalizedxcenter = xcenter / WIDTH normalizedycenter = ycenter / HEIGHT normalizedwidth = width / WIDTH normalizedheight = height / HEIGHT

Create the annotation string in YOLO format

content = f"{classid} {normalizedxcenter} {normalizedycenter} {normalizedwidth} {normalized_height}" print(content) The above conversion will give us YOLO format string. txt 0 0.46875 0.3125 0.3125 0.3125 ```

Troubleshooting [documentation page]

• Ensure that both the client and server are running.
• Check the browser console and terminal for any errors and troubleshoot accordingly.
• Verify that dependencies are correctly installed.

Contributing

If you would like to contribute to this project, please fork the repository and submit a pull request. For major changes, open an issue first to discuss your proposed changes. Additionally, please adhere to the code of conduct. More information about contributing can be found here.

License

This project is licensed under the MIT License.

Reporting Security Issues

If you find a security vulnerability in annotate-lab, please read our Security Policy for instructions on how to report it securely.

Acknowledgment

This project is detached from idapgroup's react-image-annotate, which is licensed under the MIT license, and it uses some work from image_annotator.