Papers › Saliency Maps Give a False Sense of Explanability to Image Classifiers: An Empirical...

Saliency Maps Give a False Sense of Explanability to Image Classifiers: An Empirical Evaluation across Methods and Metrics

3 Dec 2024ACML 2024 12archive 2025-07-28

Hanwei Zhang, Felipe Torres Figueroa, Holger Hermanns

The interpretability of deep neural networks (DNNs) has emerged as a crucial area of research, particularly in image classification tasks where decisions often lack transparency. Saliency maps have been widely used as a tool to decode the inner workings of these networks by highlighting regions of input images deemed most influential in the classification process. However, recent studies have revealed significant limitations and inconsistencies in the utility of saliency maps as explanations. This paper aims to systematically assess the shortcomings of saliency maps and explore alternative approaches to achieve more reliable and interpretable explanations for image classification models. We carry out a series of experiments to show that 1) the existing evaluation does not provide a fair nor meaningful comparison to the existing saliency maps; these evaluations have their implicit assumption and are not differentiable; 2) the saliency maps do not provide enough information on explaining the accuracy of network, the relationship between classes and the modification of the images.

PaperPDFCode

Syntology This paper has no arXiv id in the archive, and this site joins papers to Syntology's graph by arXiv id, so this page does not say whether the graph holds it.

Code

ftorres11/saliencysense pytorchMITSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to RISE: Randomized Input Sampling for Explanation of Black-box Models (arXiv:1806.07421; that paper's own run record: 32 ran (of which 0 constructed an object rather than computing a result; 32 with no instrument failure: 1 honoured, 0 violated, 31 with no contract checked; 0 where Syntology's instrument failed) · 2 unverified); nothing checks that they implement this paper's method. report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

A paper named beside a repository with no sample linked to this paper is shown with that paper's own run record, not this paper's: “ran” means executed on a synthesized input, not that the code is correct, and “where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Papers are listed in arXiv-id order, at most three per repository.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Image Classificationimage-classification

Results from the paper archive 2025-07-28

No leaderboard rows for this paper in the archive.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections