frozencherry/Forgery-Localization-App
1
Forgery Localization App
This Hugging Face Space hosts a Gradio application for detecting forged regions in images using an ELA (Error Level Analysis) and CNN based model.
How to use
- Upload an image (JPG, PNG, TIF).
- The application will process the image and display a forgery mask, highlighting potential manipulated areas.
Model Details
The core of this application is a TensorFlow/Keras model (final_finetune.keras) that takes two inputs: the original image and its ELA preprocessed version. The model outputs a single-channel mask indicating forged regions.
Local Development
To run this application locally, ensure you have Python 3.9+ and the following dependencies installed in a virtual environment:
pip install -r requirements.txtThen, run the application:
python app.pyThe Gradio interface will be accessible via your web browser.
