Explanations for Occluded Images
This addresses a specific limitation in interpretability for image classification, particularly in real-world scenarios with occlusions, representing an incremental improvement over existing tools.
The paper tackles the problem of poor performance of existing image classifier explanation algorithms on partially occluded images by introducing a novel black-box algorithm based on causal theory, implemented in DEEPCOVER, which achieves much more accurate explanations for occluded images and comparable performance to state-of-the-art on non-occluded images.
Existing algorithms for explaining the output of image classifiers perform poorly on inputs where the object of interest is partially occluded. We present a novel, black-box algorithm for computing explanations that uses a principled approach based on causal theory. We have implemented the method in the DEEPCOVER tool. We obtain explanations that are much more accurate than those generated by the existing explanation tools on images with occlusions and observe a level of performance comparable to the state of the art when explaining images without occlusions.