|
|
--- |
|
|
title: Deepfake Explainer App |
|
|
emoji: 🔍 |
|
|
colorFrom: purple |
|
|
colorTo: blue |
|
|
sdk: streamlit |
|
|
sdk_version: 1.32.0 |
|
|
app_file: app.py |
|
|
pinned: false |
|
|
--- |
|
|
|
|
|
# Deepfake Image Analyzer |
|
|
|
|
|
This Streamlit app uses a fine-tuned Llama 3.2 Vision model to analyze images for signs of deepfake manipulation. |
|
|
|
|
|
## Features |
|
|
|
|
|
- Upload any image to analyze for deepfake indicators |
|
|
- Get both technical and non-technical explanations |
|
|
- Customize the analysis with your own instructions |
|
|
- Adjust generation parameters like temperature and response length |
|
|
|
|
|
## Model Details |
|
|
|
|
|
This app uses the [saakshigupta/deepfake-explainer-1](https://huggingface.co/saakshigupta/deepfake-explainer-1) model, which is a fine-tuned version of Llama 3.2 Vision. |
|
|
|
|
|
## Usage Tips |
|
|
|
|
|
1. Upload a clear image of a face |
|
|
2. The model works best with front-facing portraits |
|
|
3. For more detailed analysis, increase the maximum response length |
|
|
4. You can customize the instruction in the sidebar |
|
|
|
|
|
## Technical Notes |
|
|
|
|
|
The model implements a special cross-attention mask fix to handle vision features properly. |
