Discriminator-Guided Adaptive Diffusion for Source-Free Test-Time Adaptation under Image Corruptions

TL;DR

This paper introduces a diffusion-based test-time adaptation method that dynamically controls noise application to improve robustness against various image corruptions without retraining source models.

Contribution

It presents a discriminator-guided adaptive diffusion strategy that determines optimal diffusion depth per image, enhancing robustness to diverse corruptions in source-free domain adaptation.

Findings

Achieves balanced robustness across 15 corruption types.

Demonstrates improved performance over existing methods.

Reveals adaptive diffusion schedule responds to corruption characteristics.

Abstract

In this work, we study Source-Free Unsupervised Domain Adaptation under corruption-induced domain shifts, where performance degradation is caused by natural image corruptions that go beyond additive noise, including blur, weather effects, and digital artifacts. We propose a diffusion-based, input-level adaptation framework that operates entirely at test time and keeps all source-trained models frozen, explicitly targeting robustness to corrupted target inputs. Our method leverages a source-trained diffusion model as a generative prior and introduces a discriminator-guided adaptive diffusion strategy that dynamically controls the amount of perturbation applied to each test sample. Rather than relying on a fixed diffusion depth, the discriminator determines, on a per-image basis, when sufficient forward diffusion has been applied to suppress corruption-specific artifacts, with each corruption type effectively defining a distinct target domain. This adaptive stopping mechanism applies only the necessary amount of noise to remove domainspecific corruption while preserving class-discriminative structure. The reverse diffusion process then reconstructs a source-aligned image, optionally stabilized through structural guidance, which is classified using a frozen source-trained classifier. We evaluate the proposed approach across a broad spectrum of corruption-induced target domains, covering 15 diverse corruption types, and demonstrate more balanced robustness with competitive or improved performance across non-noise corruptions. Additional analyses reveal how the adaptive diffusion schedule responds to different corruption characteristics, highlighting the practicality, generality, and robustness of the proposed framework. The code is publicly available at https://github.com/fmolivato/dgadiffusion/.