AI Paper Atlas

Research map / 3D vision / 3D generation

Bootstrap3D: Improving Multi-view Diffusion Model with Synthetic Data

Zeyi Sun, Tong Wu, Pan Zhang, Yuhang Zang, Xiaoyi Dong, Yuanjun Xiong, Dahua Lin, Jiaqi Wang

ICCV 2025 · 3D vision › 3D generation · approach: Score distillation for text-to-3D · 10 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 100k synthetic multi-view images without TTR (ablation)Cap3D-captioned multi-view diffusion · Generated 3D objectsFID (PG2.5): 120.4 -> 134.6Hurt
add 100k synthetic images with TTR (ablation)Cap3D-captioned multi-view diffusion · Multi-view imagesFID (PG2.5): 101.3 -> 60.8Improved
add 100k synthetic images with TTR (ablation)Cap3D-captioned multi-view diffusion · Multi-view imagesCLIP-R Score: 77.9 -> 83.3Improved
add training timestep reschedule to synthetic data (ablation)Cap3D + 100k synthetic · Generated 3D objectsCLIP-R Score: 71.2 -> 80.2Improved
add training timestep reschedule to synthetic data (ablation)Cap3D + 100k synthetic · Multi-view imagesFID (PG2.5): 92.0 -> 60.8Improved
replace Cap3D captions with MV-LLaVA dense recaptions (ablation)Bootstrap3D + 100k synthetic · Multi-view imagesCLIP-R Score: 83.3 -> 87.4Improved
replace Cap3D captions with MV-LLaVA dense recaptions (ablation)Bootstrap3D + 100k synthetic · Multi-view imagesFID (PG2.5): 60.8 -> 50.2Improved
increase synthetic data from 100k to 500k (ablation)Bootstrap3D · Multi-view imagesFID (PG2.5): 50.2 -> 42.4Improved
increase synthetic data from 100k to 500k (ablation)Bootstrap3D · Multi-view imagesCLIP-R Score: 87.4 -> 88.8Improved
increase synthetic data from 100k to 500k (ablation)Bootstrap3D · Generated 3D objectsCLIP-R Score: 85.1 -> 86.3Improved
raise TTR threshold T for synthetic data (ablation)Bootstrap3D · Multi-view imagesview consistency vs prompt following: T=200 chosenMixed
use Bootstrap3D instead of Instant3D (unofficial)text-to-multi-view · Real user promptsCLIP-R Score: 77.0 -> 83.5Improved

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