| title | DeepFake Detector |
|---|---|
| emoji | 🕵️ |
| colorFrom | gray |
| colorTo | red |
| sdk | docker |
| pinned | false |
A professional-grade forensic tool designed to detect manipulation in digital media. This system not only identifies if a video is fake but specifically identifies where (temporally) the manipulation occurs.
Launch App on Hugging Face Spaces
- Multi-Modal Analysis: Supports both Video (MP4, AVI, MOV) and Image (JPG, PNG) inputs.
- Temporal Localization: Instead of a single "Fake" label, we provide a timeline showing exactly which seconds of a video are manipulated.
- Forensic Dashboard:
- Interactive Video Player with "Red Zone" navigation.
- Confidence scoring per segment.
- JSON Export for forensic reporting.
- Privacy First: No data is permanently stored. Files are processed in memory and wiped effectively after analysis.
The system is built on a decoupled, scalable architecture:
- Frontend: React 19 + Vite + TailwindCSS (Single Page Application).
- Backend: FastAPI (Python) with asyncio for high-concurrency processing.
- Engine: OpenCV for frame extraction + PyTorch (Ready) for inference.
- Deployment: Dockerized multi-stage build (Node.js -> Python) on Hugging Face Spaces.
Note: This deployment is currently running in Logic Verification Mode.
- The system uses a Mock Expert engine to demonstrate the pipeline, UI, and reporting capabilities.
- It generates simulated detection scores to fully validate the frontend interaction and backend orchestration.
- Real Model Integration: The architecture is "Plug-and-Play" ready for the final trained model weights.
- Clone the repo:
git clone https://github.com/YourUsername/DeepFakeDetector.git cd DeepFakeDetector - Run with Docker (Recommended):
docker build -t deepfake-detector . docker run -p 7860:7860 deepfake-detector - Access: Open
http://localhost:7860.
Developed for the DeepFake Detection Architecture Challenge.