Interpretable and Explainable Deep Learning for Image Processing

8 minute read


Interpretability, explainability, or the lack of either, have been popular talking points in the deep learning community. The terms are generally ill-defined and the black-box nature of models, such as deep neural networks, makes it challenging to harness their power in real world applications. In this blog post, I explore notions of interpretability and explainability for both humans and machines, and discuss how we can even begin to make progress towards truly interpretable or explainable algorithms for image processing.