evaluation_kit/requirements.txt

# Only needed for evaluate_from_hf.py (loading RLALT/ACoPPer or RLALT/ACoPDoc
# from the Hugging Face Hub). Every other script in this kit — including
# convert_predictions_to_evaluation_csv.py and everything under evaluation/
# and box_grouping/ — is pure standard library, no dependencies required.
datasets>=2.14
pillow>=10.0