AI Paper Atlas

Research map / Image generation / Editing and style transfer

Realistic Face Reconstruction from Facial Embeddings via Diffusion Models

Dong Han, Yong Li, Joachim Denzler

AAAI 2026 · Image generation › Editing and style transfer · approach: Diffusion face generation and editing · 0 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
replace MLP mapping with KAN mapping in FEM (ablation)FEM · CelebA-HQ, IR152ASR average (%): 71.4 -> 77.2Improved
replace MLP mapping with KAN mapping in FEM (ablation)FEM · CelebA-HQ, IRSE50ASR average (%): 81.5 -> 83.7Improved
replace MLP mapping with KAN mapping in FEM (ablation)FEM · CelebA-HQ, MinusFaceASR average (%): 67.1 -> 67.6No clear effect
replace MLP mapping with KAN mapping in FEM (ablation)FEM · LADN-M makeupASR average (%): 81.1 -> 85.1Improved
add the FEM embedding mapping before IPA-FaceID (ablation)IPA-FaceID · CelebA-HQ, IRSE50ASR average (%): 7.5 -> 83.7Improved
apply makeup to the face images used for embeddingsFaceTI · LADNASR average (%): 74.5 -> 56.4Hurt
apply makeup to the face images used for embeddingsMAP2V · LADNASR average (%): 84.3 -> 75.5Hurt
apply makeup to the face images used for embeddingsFEM-KAN · LADNASR average (%): 91.5 -> 85.1Hurt
use FEM instead of MAP2V on HFCF PPFR embeddingsFEM-KAN · CelebA-HQ, HFCFASR average (%): 28.0 -> 83.1Improved
reduce leaked embedding fraction (partial leakage) for reconstructionFEM-KAN · IRSE50 partial embeddings, ArcFaceASR (%): 32.5% ASR below 30% leakageHurt
cloak images with Fawkes before extracting embeddingsFEM-KAN · CelebA-HQ IRSE50, ElasticFaceASR (%): 93.8 -> 47.1Hurt
use FEM instead of FaceTI for trainingmapping training cost · 90% FFHQtraining time per epoch: 51 hrs -> 3 hrsImproved

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