GuardianAI: Detect, Blur, and Organize Nude/Semi-Nude Images

This Python project detects nude and semi-nude images using the OpenNSFW2 library, blurs them, and organizes them into separate folders. It is designed to help automate the process of identifying and handling sensitive content in image datasets.

Table of Contents

1. Features
2. Prerequisites
3. Installation
4. Usage
5. Folder Structure
6. Customization
7. Troubleshooting
8. Contributing
9. License

---

Features

• Detect Nude/Semi-Nude Images: Uses the OpenNSFW2 model to classify images based on their NSFW probability.
• Blur Unsafe Images: Applies a Gaussian blur to detected nude/semi-nude images to anonymize sensitive content.
• Organize Images:
  • Moves original unsafe images to a dedicated folder (nude_images).
  • Moves blurred versions of unsafe images to another folder (blurred_images).
• Threshold-Based Filtering: Allows you to adjust the sensitivity of detection using a configurable threshold.

---

Prerequisites

Before running the application, ensure you have the following installed:

• Python 3.7 or higher
• opennsfw2 (for NSFW detection)
• Pillow (for image processing)
• shutil (for file operations)

You can install the required dependencies using the following command:

pip install opennsfw2 pillow

---

Installation

1. Clone this repository:

   git clone https://github.com/wprashed/guardianai
   cd guardianai

2. Install the required dependencies:

   pip install -r requirements.txt

3. Create the necessary folders:

   mkdir input_images nude_images blurred_images
   

4. Place your images in the input_images folder.

---

Usage

1. Run the script:

   python app.py

2. The script will:

   • Process all images in the input_images folder.
   • Move detected nude/semi-nude images to the nude_images folder.
   • Blur the detected images and move the blurred versions to the blurred_images folder.

3. Adjust the sensitivity of detection by modifying the threshold variable in the script:

   threshold = 0.6  # Default value; adjust between 0 and 1

---

Folder Structure

The project uses the following folder structure:

project/
├── input_images/       # Place your images here
├── nude_images/        # Original nude/semi-nude images will be moved here
├── blurred_images/     # Blurred nude/semi-nude images will be moved here
├── temp_blurred/       # Temporary folder for blurred images (auto-cleaned)
├── app.py              # The Python script
├── README.md           # Project documentation
└── requirements.txt    # List of dependencies

---

Customization

1. Adjust Blur Intensity: Modify the radius parameter in the blur_image() function to control the strength of the blur:

   blurred_img = img.filter(ImageFilter.GaussianBlur(radius=20))  # Default radius

2. Change Detection Threshold: Adjust the threshold variable in the process_images() function to make the detection stricter or more lenient:

   threshold = 0.6  # Higher values make detection stricter

3. Output Folder Names: Change the folder names in the process_images() function if you want to use different folder names:

   nude_folder = "nude_images"
   blurred_folder = "blurred_images"

---

Troubleshooting

1. Error: Pillow Not Installed: If you encounter an error related to Pillow, install it using:

   pip install pillow

2. Error: File Overwrite Issues: Ensure filenames are unique or modify the script to append a suffix to filenames when saving blurred images.

3. Temporary Folder Cleanup: The temp_blurred folder is automatically deleted if empty. If issues persist, manually delete it after running the script.

---

Contributing

Contributions are welcome! If you'd like to contribute, please follow these steps:

1. Fork the repository.
2. Create a new branch for your feature or bug fix.
3. Submit a pull request with a clear description of your changes.

---

License

This project is licensed under the MIT License. Feel free to use, modify, and distribute it as per the terms of the license.

---

Acknowledgments

• Thanks to the creators of the OpenNSFW2 library for making it publicly available.
• Special thanks to the Pillow library for providing robust image processing capabilities.