Align-Deform-Subtract: An Interventional Framework for Explaining Object Differences
Abstract
Given two object images, how can we explain their differences in terms of the underlying object properties? To address this question, we propose Align-Deform-Subtract (ADS)---an interventional framework for explaining object differences. By leveraging semantic alignments in image-space as counterfactual interventions on the underlying object properties, ADS iteratively quantifies and removes differences in object properties. The result is a set of "disentangled" error measures which explain object differences in terms of their underlying properties. Experiments on real and synthetic data illustrate the efficacy of the framework.
Cite
Text
Eastwood et al. "Align-Deform-Subtract: An Interventional Framework for Explaining Object Differences." ICLR 2022 Workshops: OSC, 2022.Markdown
[Eastwood et al. "Align-Deform-Subtract: An Interventional Framework for Explaining Object Differences." ICLR 2022 Workshops: OSC, 2022.](https://mlanthology.org/iclrw/2022/eastwood2022iclrw-aligndeformsubtract/)BibTeX
@inproceedings{eastwood2022iclrw-aligndeformsubtract,
title = {{Align-Deform-Subtract: An Interventional Framework for Explaining Object Differences}},
author = {Eastwood, Cian and Nanbo, Li and Williams, Chris},
booktitle = {ICLR 2022 Workshops: OSC},
year = {2022},
url = {https://mlanthology.org/iclrw/2022/eastwood2022iclrw-aligndeformsubtract/}
}