🚨 Lean Guard – AI-Based Image Moderator

Lean Guard is a powerful AI-driven image moderation system designed to automatically classify uploaded images as Safe or Prohibited. Leveraging deep learning models, it detects sensitive content such as weapons, gore, and explosives — helping platforms maintain safe and compliant environments.

---

⚙️ Tech Stack

• Frontend: HTML5, CSS3 (via Flask templating – Jinja2)
• Backend: Python 3.10, Flask (micro web framework)
• Machine Learning: Keras, TensorFlow, EfficientNet, Custom CNN
• Deployment: Localhost (Flask server)
• Environment: venv (Python virtual environment)
• Other Tools: Git, GitHub, Google Colab (for model training and prototyping)

🔍 Features

• 🌐 Interactive Web Interface for uploading and analyzing images in real-time
• 🧠 Deep Learning Models trained on sensitive content categories
• ⚡ Instant Feedback on whether an image is safe or violates content policy
• 🔒 Extendable Architecture for future authentication, analytics, or cloud integrations

---

🧠 Models Used

The following Keras-based models are integrated into the backend for prediction:

• best_weapon_model.keras – Detects the presence of weapons
• best_gore_model.keras – Identifies gory or violent imagery
• best_explosive_model.keras – Recognizes explosives
• custom_cnn_model.keras – Custom-built CNN for benchmarking
• efficientnet_model.keras – Transfer learning using EfficientNet for high accuracy

---

📁 Folder Structure

LEAN_GUARD/
│
├── images_for_testing/       # Sample images for testing
│   ├── safe_samples/         # Approved content examples
│   └── prohibited_samples/   # Restricted content examples
│
├── models/                   # Pretrained Keras models
│   ├── best_explosive_model.keras
│   ├── best_gore_model.keras
│   ├── best_weapon_model.keras
│   ├── custom_cnn_model.keras
│   └── efficientnet_model.keras
│
├── screenshots/              # Application screenshots
│   ├── home.jpg
│   ├── safe_image_result.jpg
│   ├── safe_image_result_2.jpg
│   ├── prohibited_result.jpg
│   └── prohibited_result_2.jpg
│
├── templates/                # Flask templates
│   └── index.html            # Main interface
│
├── venv/                     # Python virtual environment
│
├── app.py                    # Flask application
├── requirements.txt          # Python dependencies
└── README.md                 # Project documentation

---

🚀 Getting Started

1️⃣ Clone the Repository

git clone https://github.com/yourusername/lean_guard.git cd LEAN_GUARD

2️⃣ Create a Virtual Environment

python -m venv venv

Activate the environment:

On Windows:

venv\Scripts\activate

On Mac/Linux:

source venv/bin/activate

3️⃣ Install Dependencies

pip install -r requirements.txt

4️⃣ Run the Flask App

python app.py 🔗 The app will be accessible at: http://127.0.0.1:8000

💡 How It Works

User uploads an image via the web interface (index.html)

The Flask backend receives and preprocesses the image

The image is passed through the following models:

• 🔫 Weapon detection

• 🩸 Gore detection

• 💣 Explosives detection

If any model flags prohibited content, the image is marked Prohibited

The result is returned to the frontend with dynamic feedback and UI updates

🧪 Testing

Use the images for testing/ folder to simulate safe and unsafe inputs. These are example images demonstrating the system’s ability to detect inappropriate content across different classes.

🛠 Custom Training (Optional)

To build and use your own models:

Prepare labeled datasets for:

Safe vs Weapon

Safe vs Gore

Safe vs Explosives

Train using Keras (CNN or EfficientNet recommended)

Save models as .keras and place them in the models/ folder

🖼️ Screenshots

🔹 Home Page

🔹 Safe Image Results

🔹 Prohibited Image Results

👥 Contributors

• Seemab Hassan

📄 License

This project is licensed under the MIT License – feel free to use and modify with credit.

📌 Future Improvements

User authentication for moderators

Logging and dashboard analytics

Multi-language support

Integration with content management systems or cloud storage