IConFace: Identity-Structure Asymmetric Conditioning for Unified Reference-Aware Face Restoration

Axi Niu; Jinyang Zhang; Senyan Qing

IConFace: Identity-Structure Asymmetric Conditioning for Unified Reference-Aware Face Restoration

Axi Niu^* Jinyang Zhang^* Senyan Qing

Northwestern Polytechnical University

nax@nwpu.edu.cn zhangjinyang@mail.nwpu.edu.cn qingsenyan@nwpu.edu.cn

^*Equal contribution

Project Paper Supplementary Code Coming Soon BibTeX

Reference 1

Reference 2

Reference 3

Degraded Input

CodeFormer

FaceMe

Ours

Ground Truth

IConFace preserves reference-consistent facial details better than strong blind and reference-aware baselines while remaining anchored to the degraded input.

CodeFormer

FaceMe

Ours

Ground Truth

Abstract

Blind face restoration is highly ill-posed under severe degradation, where identity-critical details may be missing from the degraded input. Same-identity references reduce this ambiguity, but mismatched pose, expression, illumination, age, makeup, or local facial states can lead to overuse of reference appearance. We propose IConFace, a unified reference-aware and no-reference framework with identity--structure asymmetric conditioning. References are distilled into a norm-weighted global AdaFace identity anchor for image-only modulation, while the degraded image is reinforced as the spatial structure anchor through low-rank residuals and block-wise degraded cross-attention with two-route memory. The resulting single checkpoint exploits references when available and falls back to no-reference restoration when absent, improving identity consistency, fine-detail recovery, and degraded-only restoration quality in a unified model.

A unified optional-reference formulation where one checkpoint supports reference-aware and no-reference restoration through asymmetric identity--structure conditioning.
Lightweight pathways that use norm-weighted AdaFace modulation for reference identity and low-rank residuals with two-route degraded memory for structure and detail anchoring.
Extensive results showing improved reference-aligned identity consistency, fine-detail recovery under severe degradation, and strong no-reference perceptual quality.

Method Overview

IConFace is built on the FLUX.2-klein-base-4B hybrid concat restoration backbone. The main sequence concatenates noisy scene tokens, degraded-image tokens, and optional reference tokens, while no-reference mode removes the reference segment rather than duplicating placeholders. On top of the backbone, a global identity pathway aggregates valid references into a norm-weighted AdaFace anchor for image-only modulation, and a degraded structure pathway reinforces the spatially aligned degraded observation with low-rank input residuals plus block-wise degraded cross-attention. The two-route memory separates base structure from local detail so the model can use reference identity evidence without copying mismatched pose, expression, illumination, age, makeup, or local facial state.

Ref 1

Ref 2

Ref 3

LQ

Ours

GT

Main Paper Figure Panels

Qualitative figure panels from the AAAI main paper, including reference-aware comparisons, no-reference comparisons, and ablation examples.

Supplementary Visual Evidence

Supplementary panels are regenerated from AAAI_IConface_supp.tex and its figure input: selected reference--GT gap diagnostics, four reference-aware datasets, five no-reference datasets, and three reference-aware ablation datasets.

BibTeX

@misc{iconface2026,
  title={IConFace: Identity-Structure Asymmetric Conditioning for Unified Reference-Aware Face Restoration},
  author={Niu, Axi and Zhang, Jinyang and Qing, Senyan},
  year={2026},
  note={Project page, paper PDF, and supplementary PDF},
  url={https://cosmicrealm.github.io/IConFace/}
}