"Let the wisdom of the ancients meet the power of artificial intelligence."
🏠 Home Interface
|
🔍 Chatbot Interaction
|
💬 Recommendation Results
|
🧠 This video ad was created entirely using AI tools.
Have You Ever Lived History? Horus AI is more than just software — it’s a digital bridge between the past and the present. It reimagines how we interact with ancient Egyptian artifacts using the power of AI.
- Egypt's Legacy: One of humanity’s greatest civilizations
- Beyond Sight: Experience history, not just see it
- Technology’s Role: Live history through advanced tech
Current AI models like ChatGPT, Claude, and Gemini often:
- Provide general or vague responses
- Misclassify historical images
- Lack deep cultural understanding of ancient artifacts
Horus AI teaches artificial intelligence to truly understand history — not just recognize it.
| Feature | Description |
|---|---|
| 📸 Image Classification | Identify ancient artifacts with CNN + transfer learning |
| 📝 Descriptions | Generate accurate and engaging historical content |
| 🗺️ Recommendations | Get personalized site suggestions and travel tips |
| 💬 Virtual Guide | Ask questions via a smart Gemini-powered chat assistant |
| Technology | Role |
|---|---|
| 🧠 Keras + TensorFlow | Image classification using transfer learning |
| 🔤 Google Gemini API | Generates context-aware responses and historical explanations |
| 🌐 Flask | User-friendly web app to access Horus AI |
| ⚙️ Modular Codebase | Efficient and maintainable project structure |
-
Curated top-quality datasets of ancient Egyptian artifacts
-
Augmentation Techniques:
- Rotation, flipping, zooming, cropping
- Brightness/contrast shifts
- Resize images
- Filter low-resolution data
- Balance underrepresented and overrepresented classes
-
Model: CNN using Keras
-
Initial Accuracy: ~50%
-
Improvements:
- Hyperparameter tuning
- Class merging and relabeling
- Manual data verification
- Final Accuracy: ~80%
- Issue: Classes like Ramessum and Ramesseum were split unnecessarily
- Fix: Merged confusing or duplicate classes
- Result: Reduced misclassification (e.g., Sphinx misclassified under Giza_Pyramid_Complex)
- 🧪 Advanced augmentations
- 🔁 More real-world data collection
- 💻 Better pretrained models
- 🧹 Clean, labeled, and balanced datasets
Horus AI Assistant is built using Google Gemini Pro:
- 💡 Provides cultural, accurate explanations
- 🗣️ Responds to historical queries interactively
- 📌 Integrated feedback loop for better personalization
| Component | Description |
|---|---|
| 🧠 Model Integration | model_utils.py handles classification logic |
| 🔤 NLP Utilities | llm_utils.py manages Gemini API interactions |
| 🌐 Frontend | Flask + HTML/CSS for uploading, viewing, and chatting |
- Upload Image: Submit an artifact image
- Classification: AI identifies the artifact
- Description: Get historical context
- Recommendations: Travel site suggestions
- Live Q&A: Chat with Horus AI
flask_project/
├── app.py # Main Flask app
├── class_labels.py # Artifact labels
├── last_model.keras # Trained CNN model
├── llm_utils.py # Gemini API logic
├── model_utils.py # Image processing
├── requirements.txt
├── static/
│ └── images/
└── templates/
└── index.html, etc.
# Clone the repo
git clone https://github.com/Nadercr7/Horus-AI-Depi
cd flask_project
# Set up environment
python3 -m venv venv
source venv/bin/activate # Windows: venv\Scripts\activate
# Install requirements
pip install -r requirements.txt
# Configure API key
echo "GEMINI_API_KEY=your_api_key_here" > .env
# Run the app
flask runVisit: http://127.0.0.1:5000
From raw data to a polished web app with 80% classification accuracy — we combined:
- Deep learning (CNN)
- Transfer learning
- NLP (Gemini)
- Error analysis and human feedback
To create intelligent, accessible archaeology tools where:
- AI becomes a historical companion
- Learning about civilizations is immersive and personalized