Smart Tag Authentication

Supporting code for the paper "Nanocatalyst-enabled physically unclonable functions as smart anticounterfeiting tags with AI-aided smartphone authentication".

Installation

pip install pipenv
pipenv install

Dataset creation

The provided data are expected to be in filename.jpg format and placed in ./data/imgs/ folder with the related annotation file, named filename.csv.

The current dataset can be provided upon request. All the contained images should should be unzipped into ./data/imgs/.

Annotate the 4 code corners and reorganize the data using the following command:

 pipenv shell
 python3 ./src/create_dataset.py

It will create a ./dataset/ folder and TRAIN, VAL and TEST subfolders with the annotated images organized as follow:

./dataset/
    filename_1/
        0/
        1/
        2/
        3/
    filename_2/
        0/
        1/
        2/
        3/
    ...
    filename_N/
        0/
        1/
        2/
        3/

PAY ATTENTION THAT CURRENTLY THE SPLITS ARE HARD-CODED IN create_dataset.py !!!!

Model Training

The model implementation is in the classifier_ord_regr folder.

The model has been implemented in model.py, while the dataloaders have been implemented in data_module.py.

Classifier Ordinal Regression

Create Data Split

Edit ./src/classifier_ord_regr/conf/conf_datasplit.yaml for creating TRAIN, VAL and TEST data splits. Then, launch:

pipenv shell
python3 ./src/classifier_ord_regr/create_data_splits.py

Training can be launched using:

python3 ./src/classifier_ord_regr/train.py