0

Visual Explanations via Iterated Integrated Attributions

Iterated Integrated Attributions (IIA) generates accurate explanation maps for vision models by iteratively integrating input images, model representations, and gradients across various tasks, datasets, and architectures.

Year
2023
Venue
visual-explanations-via-iterated-integrated
Authors
5
Hosting
Abstract onlyARXIV-DEFAULT

Cite

Notes

Only stored in your browser.

Attribution

Abstract & full text
arxiv.org/abs/2310.18585ARXIV-DEFAULT
TL;DR
Semantic Scholar
Attribution policy →

Abstract

We introduce Iterated Integrated Attributions (IIA) - a generic method for explaining the predictions of vision models. IIA employs iterative integration across the input image, the internal representations generated by the model, and their gradients, yielding precise and focused explanation maps. We demonstrate the effectiveness of IIA through comprehensive evaluations across various tasks, datasets, and network architectures. Our results showcase that IIA produces accurate explanation maps, outperforming other state-of-the-art explanation techniques.

Authors

5