The main purpose of this project is to distinguish between AI-Generated and Authentic photos.
This GitHub repository is structured for presentation purposes and showcases six models that were trained and introduced in the accompanying report.
-
checkpoints
Contains saved model weights. Due to file size, some models are provided as google drive links. -
graphics
Includes visualizations and graphics for all the models. -
metrics
Stores the evaluation metrics for all the models. -
data
Folder where we loaded the data
-
models
Contains the implementation of all the models. -
utils
Houses support functions such as:- Dataloaders for every model separately
- Early stopping
- Evaluation function
-
training.ipynb
A Jupyter notebook with the training loop for one of the models.