Python AI Image Captioning

This project provides a practical set of tools for generating captions for images using the BLIP model. You can caption images by uploading a single file, processing all images in a local folder, or scraping images from a webpage. The interface is simple and user-friendly, making it easy for anyone to use.

Table of Contents

• Features
• Screenshots
• Installation
• Usage
• Example Output
• Notes
• Contributing & License
• Acknowledgements

Features

• Single Image Upload: Instantly get a caption for any image you upload via the web interface.
• Bulk Local Captioning: Automatically process all images in a folder and save their captions to a file.
• Webpage Scraping: Find and caption all images from any webpage you provide.
• Easy Web Interface: All tools use Gradio-based web UIs for convenience.
• Docker Support: Run the whole project easily in a container.

Screenshots

Single Image Upload:

---

Bulk Local Captioning:

---

Webpage Scraping:

---

Installation

1. Clone the repository:

   git clone https://github.com/eray-yuztyurk/python-ai-image-captioning.git
   cd python-ai-image-captioning

2. Make sure Python 3.10+ is installed.

3. Create and activate a virtual environment:

   python3.10 -m venv venv
   source venv/bin/activate

4. Install dependencies:

   pip install --upgrade pip
   pip install -r requirements.txt

Usage

Start All Interfaces Together

python3.10 main.py

Each interface will open on its own port (7860, 7861, 7862). If your browser does not open automatically, visit these addresses manually.

Run Individual Scripts

• Single image upload:
  python3.10 uploaded_image_captioner.py
  (Port: 7860)

• Local folder images:
  python3.10 local_img_captioner_automated.py
  (Port: 7862)

• Webpage scraping:
  python3.10 url_img_captioner_automated.py
  (Port: 7861)

Run with Docker

docker build -t ai-image-captioning .
docker run -p 7860:7860 -p 7861:7861 -p 7862:7862 ai-image-captioning

Example Output

cat.jpg: a small orange cat sitting on a windowsill
https://example.com/image1.png: a group of people standing in front of a building

Notes

• Make sure the outputs/ directory exists, or specify a valid output path.
• The BLIP model will be downloaded automatically on first run.
• For best results, use clear and sufficiently large images (at least 100x100 pixels).

Contributing & License

This project is released under the MIT License. If you want to contribute, feel free to open an issue or pull request.

Acknowledgements

• Salesforce BLIP
• Gradio
• HuggingFace Transformers

Feel free to open issues or contribute improvements!