hectorandac/craft-sentence-extractor repository overview

⁠CRAFT-Sentence-Extractor

CRAFT-Sentence-Extractor is a Docker image designed for extracting sentences from documents using the CRAFT (Character Region Awareness for Text Detection) model, enhanced with custom logic for sentence grouping and document layout understanding.

⁠Key Features:

• Text Detection: Detects words in documents using the CRAFT model.
• Sentence Grouping: Groups words into sentences based on spatial relationships.
• Flexible Configuration: Handles multi-column layouts with customizable hyperparameters.
• Future Ready: Designed to integrate with Transformer models for advanced document layout analysis.

⁠How to Use

To run the Docker container and expose it on port 5000:

docker run -d -p 5000:5000 your-dockerhub-username/craft-app

Once the container is running, you can send an image (e.g., a scanned book page) via an HTTP request to the /process_image endpoint.

Example:

curl -X POST -F "image=@your_image.png" http://localhost:5000/process_image

This will return detected sentences or word bounding boxes, depending on the configuration.

⁠Environment Variables

You can set the following environment variables to adjust the app’s behavior:

• FLASK_ENV: Set to production (default) or development.

⁠License

This project is licensed under the MIT License.

⁠Contributions

Contributions and suggestions are welcome! Feel free to open an issue or submit a pull request.