Explainable machine learning in image classification models: An uncertainty quantification perspective