Argus

Argus is an AI-generated image detection research project focused on reproducible binary classification (Real vs. AI-generated) with PyTorch baselines and scaffolding for probing frozen JEPA representations.

Setup

• Python 3.10+
• Install dependencies:

pip install -r requirements.txt

Data layout

Point training to a dataset directory with this layout:

dataset/
  real/
    image_1.jpg
    ...
  fake/
    image_1.jpg
    ...

Train baseline

python train.py --data-dir /absolute/path/to/dataset