AI Paper Atlas

Research map / Image generation / Editing and style transfer

UniPortrait: A Unified Framework for Identity-Preserving Single- and Multi-Human Image Personalization

Junjie He, Yifeng Geng, Liefeng Bo

ICCV 2025 · Image generation › Editing and style transfer · approach: Diffusion face generation and editing · 45 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 penultimate-layer local face features instead of global feature (ablation)UniPortrait ID embedding · Single-ID ablation of ID embedding moduleFace similarity (%): 45.5 -> 58.3Improved
add CLIP local face-structure features (ablation)UniPortrait ID embedding · Single-ID ablation of ID embedding moduleFace similarity (%): 58.3 -> 61.4Improved
add face backbone shallow features to structure branch (ablation)UniPortrait ID embedding · Single-ID ablation of ID embedding moduleFace similarity (%): 61.4 -> 64.5Improved
add face structure branch (ablation)UniPortrait ID embedding · Single-ID ablation of ID embedding moduleFID: 120.3 -> 124.7Hurt
add DropToken and DropPath on face structure branch (ablation)UniPortrait ID embedding · Single-ID ablation of ID embedding moduleFace similarity (%): 64.5 -> 71.1Improved
add DropToken and DropPath on face structure branch (ablation)UniPortrait ID embedding · Single-ID ablation of ID embedding moduleFID: 128.1 -> 123.4Improved
add DropToken and DropPath on face structure branch (ablation)UniPortrait ID embedding · Single-ID ablation of ID embedding moduleCLIP-T (%): 25.9 -> 26.1No clear effect
add face structure branch (local feature only vs full) (ablation)UniPortrait ID embedding · Single-ID ablation of ID embedding moduleCLIP-T (%): 26.8 -> 26.1Hurt
add routing regularization loss (ablation)UniPortrait ID routing · Multi-ID (105 pairs)Face similarity (%): 59.1 -> 67.3Improved
add routing regularization loss (ablation)UniPortrait ID routing · Multi-ID (105 pairs)CLIP-T (%): 27.5 -> 27.4No clear effect
use 4 reference images instead of 1 (ablation)UniPortrait · Single-ID multi-referenceFace similarity (%): 55.5 -> 61.2Improved
use 4 reference images instead of 1 (ablation)UniPortrait · Single-ID multi-referenceCLIP-T (%): 26.1 -> 25.9No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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