Update README.md
Browse files
README.md
CHANGED
|
@@ -1,14 +1,64 @@
|
|
| 1 |
---
|
| 2 |
title: DEEP FAKE DETECTOR
|
| 3 |
-
emoji:
|
| 4 |
-
colorFrom:
|
| 5 |
-
colorTo:
|
| 6 |
sdk: gradio
|
| 7 |
-
sdk_version:
|
| 8 |
app_file: app.py
|
| 9 |
-
pinned:
|
| 10 |
-
license:
|
| 11 |
-
|
| 12 |
---
|
| 13 |
|
| 14 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
title: DEEP FAKE DETECTOR
|
| 3 |
+
emoji: π
|
| 4 |
+
colorFrom: blue
|
| 5 |
+
colorTo: purple
|
| 6 |
sdk: gradio
|
| 7 |
+
sdk_version: 4.36.0
|
| 8 |
app_file: app.py
|
| 9 |
+
pinned: true
|
| 10 |
+
license: mit
|
| 11 |
+
short_description: Multimodal deepfake detection across 5 modalities
|
| 12 |
---
|
| 13 |
|
| 14 |
+
# Multimodal Deepfake Detection System
|
| 15 |
+
|
| 16 |
+
A modular AI system for detecting fake content across 5 modalities β text, audio, image, video, and lip-sync β using hybrid CNN-Transformer-LLM architectures.
|
| 17 |
+
|
| 18 |
+
## Live Demo
|
| 19 |
+
|
| 20 |
+
Try the live system using the tabs above:
|
| 21 |
+
|
| 22 |
+
- **Text Verification** β Detects fake news using SERP API + DeBERTa + Llama-3.3
|
| 23 |
+
- **Audio Forensics** β Identifies AI-cloned voices using Wav2Vec2 + AST + CLAP
|
| 24 |
+
- **Image Analysis** β Detects deepfake images using ResNet-18 + Vision Transformer
|
| 25 |
+
- **Video Forensics** β Analyzes video deepfakes with multi-frame voting + OpenCLIP
|
| 26 |
+
- **Multi-Modal Lip-Sync** β Detects audio-video mismatch using LSE-Net + Claude reasoning
|
| 27 |
+
|
| 28 |
+
## Performance
|
| 29 |
+
|
| 30 |
+
| Modality | Accuracy |
|
| 31 |
+
|----------|----------|
|
| 32 |
+
| Text | 99% |
|
| 33 |
+
| Image | 92% |
|
| 34 |
+
| Audio | 91% |
|
| 35 |
+
| Video | 91% |
|
| 36 |
+
| Lip-Sync | 89% |
|
| 37 |
+
|
| 38 |
+
## Tech Stack
|
| 39 |
+
|
| 40 |
+
- **Frameworks:** PyTorch, Hugging Face Transformers, Gradio
|
| 41 |
+
- **Vision Models:** Vision Transformer (ViT), ResNet-18, OpenCLIP
|
| 42 |
+
- **Language Models:** DeBERTa-v3, Llama-3.3, Claude 3.5 Sonnet
|
| 43 |
+
- **Audio Models:** Wav2Vec2, AST, CLAP, Whisper
|
| 44 |
+
- **Datasets:** FaceForensics++, FakeAVCeleb, ASVspoof
|
| 45 |
+
|
| 46 |
+
## About
|
| 47 |
+
|
| 48 |
+
This system was developed as part of MS thesis research at GIFT University, Pakistan.
|
| 49 |
+
|
| 50 |
+
**Author:** Aneela Pervez
|
| 51 |
+
**Institution:** GIFT University, Department of Computer Science
|
| 52 |
+
**Supervisor:** Dr. Muhammad Ziad Nayyer
|
| 53 |
+
|
| 54 |
+
## Links
|
| 55 |
+
|
| 56 |
+
- **GitHub Repository:** https://github.com/ANEELA-PERVEZ
|
| 57 |
+
- **LinkedIn:** https://linkedin.com/in/aneela-pervez-95472b310
|
| 58 |
+
- **Paper:** Coming soon on arXiv
|
| 59 |
+
|
| 60 |
+
## Citation
|
| 61 |
+
|
| 62 |
+
If you use this system in your research, please cite:
|
| 63 |
+
|
| 64 |
+
```
|