This dataset is the same as the one in
https://huggingface.co/datasets/CSE472-blanket-challenge/final-dataset.
The only difference is that test split has ground truth labels and metadata.
train = load_dataset("CSE472-blanket-challenge/final-ground-truth", split='train')
test = load_dataset("CSE472-blanket-challenge/final-ground-truth", split='test')