saliency methods are a popular class of feature attribution explanation
methods that aim to capture a model's predictive reasoning by identifying
"important" pixels in an input image. However, the development and adoption of
these methods are hindered by the lack of access to ground-tr