#🕵️‍♂️ FaceTective---Face-Authenticity-Detector FaceTective is an AI-powered tool designed to detect deepfake images with high accuracy. By leveraging MobileNetV2 and fine-tuning with custom layers, it classifies whether an uploaded face image is real or fake. The project also features a Streamlit-based web app for easy interaction.

🚀 Features

1. Detects real vs fake images using deep learning.
2. Built with MobileNetV2 as the backbone + custom classification layers.
3. Includes data augmentation (rotation, zoom, flipping) for better generalization.
4. Provides probability scores along with predictions.
5. Simple Streamlit web interface for uploading and testing images.

🛠️ Tech Stack

1. Python
2. TensorFlow / Keras – Model training & fine-tuning
3. MobileNetV2 – Pretrained CNN architecture
4. NumPy & Matplotlib – Data handling & visualization
5. Streamlit – Web app deployment

📂 Project Structure FaceTective/ │── facetective_train.py # Model training script
│── app.py # Streamlit web app
│── facetective_model.h5 # Saved trained model
│── data/ # Dataset (real & fake images)
│── README.md # Project documentation

⚙️ Installation Clone the repository: git clone https://github.com/your-username/FaceTective.git cd FaceTective

Install dependencies: pip install -r requirements.txt

Run the training script (optional, if you want to retrain): python facetective_train.py

Launch the Streamlit app: streamlit run app.py

🎯 Usage

1. Upload a face image (.jpg, .jpeg, .png).
2. Model preprocesses and classifies the image.
3. Get prediction result: Real or Fake + confidence score.

📊 Model Performance

1. Architecture: MobileNetV2 + Dense layers + Dropout
2. Loss Function: Binary Crossentropy
3. Optimizer: Adam
4. Metrics Tracked: Accuracy, Precision, Recall, F1-score, ROC-AUC

🔮 Future Improvements

1. Add explainability (Grad-CAM heatmaps).
2. Support video deepfake detection.
3. Improve dataset size and diversity.
4. Explore Vision Transformers (ViT) or XceptionNet for better results.

📸 Demo Screenshot

🤝 Contributing Pull requests are welcome! For major changes, please open an issue first to discuss what you’d like to change.