Proofie provides state-of-the-art deepfake detection technology, combining advanced deep learning models and custom AI algorithms to ensure content authenticity. Our mission is to empower developers, researchers, and organizations with scalable, high-precision tools to combat digital fraud and identity manipulation.
This repository offers access to the Prooflie API, allowing you to upload images, analyze deepfakes, and retrieve processed results through a secure, developer-friendly environment.
🔗 Live API Documentation: API Docs
- Deepfake Detection: Upload images and receive real-time deepfake analysis using an ensemble of AI models trained on high-quality datasets.
- Session Tracking: Retrieve session details, including status, face detection results, and probability scores.
- Model Accuracy: Our ensemble learning approach delivers an AUC score of 99.16%, ensuring superior performance.
- Secure & Scalable: Designed for fraud prevention, identity verification, and media integrity protection.
Send an image to the API for analysis and receive a session UUID.
Use the session UUID to get detection results, including deepfake probability scores from multiple AI models.
Retrieve the original image, detected face crops, and classification results for further analysis.
🔗 Full API Reference: Swagger Docs
Our deepfake detection system is powered by state-of-the-art deep learning models, including:
- Pre-trained Architectures: Leveraging Xception and custom CNNs fine-tuned for facial forgery detection.
- Robust Data Processing: Images undergo augmentation, normalization, and random erasing for enhanced model resilience.
- Cross-Validation: Stratified K-Fold validation ensures stable, unbiased performance.
- Ensemble Learning: Combining multiple models to improve accuracy and generalization.
🔬 Performance Metrics:
- Accuracy: 95.40%
- F1 Score: 95.42%
- Precision: 94.19%
- Recall: 96.69%
- AUC Score: 99.16%
For a deep dive into the technology, see our Research Paper.
We welcome contributions from the developer and research community! If you have ideas, bug reports, or suggestions:
- Submit a GitHub Issue
- Open a Pull Request
- Share your thoughts in Discussions
- 🌐 Website: prooflie.com
- 📖 Research & Papers: prooflie.com/research
- 🐦 X: @ProoflyAI
- 💼 LinkedIn: Prooflie