AI Paper Atlas

Research map / 3D vision / Face and shape

UP2You: Fast Reconstruction of Yourself from Unconstrained Photo Collections

Zeyu Cai, Ziyang Li, Xiaoben Li, Boqian Li, Zeyu Wang, Zhenyu Zhang, Yuliang Xiu

ICLR 2026 · 3D vision › Face and shape · approach: Full-body avatar reconstruction

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 PCFA correlation top-k aggregation instead of mean-averaging reference features (ablation)UP2You · PuzzleIOIPSNR: 17.412 -> 23.896Improved
use PCFA instead of concatenating reference features (ablation)UP2You · 4D-Dress, 12 refsPSNR / GPU memory: PSNR 23.366 -> 25.848; GPU 37.96 -> 20.88Improved
use correlation top-k selection instead of weighted sum (ablation)UP2You · PuzzleIOIPSNR: 20.167 -> 23.896Improved
use ReferenceNet instead of CLIP as reference encoder (ablation)UP2You · PuzzleIOIPSNR: 20.152 -> 23.896Improved
use ReferenceNet instead of DINOv2 as reference encoder (ablation)UP2You · PuzzleIOIPSNR: 19.744 -> 23.896Improved
use DINOv2 instead of CLIP or DINOv1 in the correlation module (ablation)PCFA · 4D-DressPSNR: 23.876 (CLIP) / 24.170 (DINOv1) -> 25.848Improved
add more reference images (3 to 12) (ablation)UP2You · 4D-DressPSNR: 24.159 -> 25.848Improved
raise top-k ratio gamma from 1.0 to 2.0 (ablation)PCFA · 4D-DressPSNR: 24.978 -> 25.848Improved
raise top-k ratio gamma from 2.0 to 3.0 (ablation)PCFA · 4D-DressPSNR: 25.848 -> 25.837No clear effect
use perceiver shape predictor instead of MLP (ablation)UP2You shape predictor · shape V2VV2V: 8.336 -> 7.399 (12 refs)Improved
use multiple references instead of single-input shape predictorUP2You shape predictor · shape V2VV2V: 9.287 (PromptHMR) -> 7.399 (12 refs)Improved
use unconstrained photos instead of single front view inputUP2You · 4D-DressPSNR (front): PSHuman 25.384 vs ours 25.257 front; back 23.382 -> 25.488Mixed

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