AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

XRAI: Better Attributions Through Regions

Andrei Kapishnikov, Tolga Bolukbasi, Fernanda Viegas, Michael Terry

ICCV 2019 · Image generation › Autoregressive and latent generation · approach: Learning-based image quality assessment · 220 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use black+white baselines instead of a black baseline for integrated gradientsIntegrated Gradients · ImageNet valarea under SIC (Resnet50-V2): 0.515 -> 0.575Improved
use black+white baselines instead of a black baseline for integrated gradientsIntegrated Gradients · ImageNet segmentation datasetAUC: 0.710 -> 0.729Improved
use 4 random baselines instead of a black baseline for integrated gradientsIntegrated Gradients · ImageNet valarea under SIC (Resnet50-V2): 0.515 -> 0.623Improved
use 4 random baselines instead of black+white baselines for integrated gradientsIntegrated Gradients · ImageNet valarea under SIC (Inception): 0.601 -> 0.595No clear effect
add region-level aggregation of IG attributions (XRAI) on top of IG B+WIG (B+W) · ImageNet valarea under SIC (Resnet50-V2): 0.575 -> 0.749Improved
add region-level aggregation of IG attributions (XRAI) on top of IG B+WIG (B+W) · ImageNet valarea under AIC (Inception): 0.634 -> 0.727Improved
use XRAI instead of GradCamsaliency method · ImageNet valarea under SIC (Inception): 0.703 -> 0.720Improved
use XRAI instead of GradCamsaliency method · ImageNet valarea under SIC (Resnet50-V2): 0.760 -> 0.749Hurt
use XRAI instead of GradCamsaliency method · ImageNet valarea under AIC (Resnet50-V2): 0.727 -> 0.728No clear effect
use XRAI instead of GradCAMsaliency method · ImageNet segmentation datasetAUC: 0.742 -> 0.836Improved
use XRAI instead of GradCAMsaliency method · ImageNet segmentation datasetMAE: 0.194 -> 0.149Improved
use XRAI instead of GradCAMsaliency method · ImageNet segmentation datasetF1: 0.715 -> 0.786Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy