AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

Realiz3D: 3D Generation Made Photorealistic via Domain-Aware Learning

Ido Sobol, Kihyuk Sohn, Yoav Blum, Egor Zakharov, Max Bluvstein, Andrea Vedaldi, Or Litany

CVPR 2026 · Image generation › Text-to-image diffusion · approach: Layout-to-image control · 1 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
add real data at Stage 2 of two-stage training (ablation)Domain Shifter model · Multiview TexturingFID_I: 212.86 -> 198.44Improved
use two-stage instead of joint training of Domain Shifters (ablation)Domain Shifter model · Multiview TexturingFID_I: 216.63 -> 198.44Improved
add domain reassignment (ablation)DS two-stage · Multiview TexturingPSNR: 23.97 -> 24.18Improved
add inference-time domain shifting (sampling) (ablation)DS two-stage · Multiview TexturingPSNR: 23.97 -> 24.25Mixed
add layer-aware training to DS + Reassign (ablation)DS + Reassign · Multiview TexturingPSNR: 24.18 -> 24.52Improved
add inference-time sampling shift to full training (ablation)DS + LA Train + Reassign · Multiview TexturingPSNR: 24.52 -> 24.78Improved
replace Syn+Real full fine-tuning with Realiz3Dinternal DiT · Multiview TexturingFID_I: 214.84 -> 200.24Improved
replace Syn Only fine-tuning with Realiz3Dinternal DiT · Multiview TexturingPSNR: 25.76 -> 24.78Hurt
replace Syn Only fine-tuning with Realiz3Dinternal DiT · Text-to-MultiviewFID_I: 215.57 -> 196.01Improved
replace Syn Only fine-tuning with Realiz3Dinternal DiT · Text-to-MultiviewPSNR: 19.66 -> 19.02Hurt
replace Domain Switcher (Joint) with Realiz3DDomain Switcher · Multiview TexturingPSNR: 23.16 -> 24.78Improved
replace LoRA rank 32 with Realiz3DLoRA · Text-to-MultiviewPSNR: 17.54 -> 19.02Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 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