Research map / 3D vision / 3D generation
Align Images Before You Generate
Shihua Zhang, Qiuhong Shen, Xinchao Wang
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add training-free CorrAdapter | MVAdapter · Image-conditioned multi-view (GSO) | MEt3R: 0.2116 -> 0.2036 | Improved |
| add training-free CorrAdapter | SyncDreamer · Image-conditioned multi-view (GSO) | depth error: 14.15 -> 12.60 | Improved |
| add training-free CorrAdapter | SyncDreamer · Image-conditioned multi-view (GSO) | Chamfer Distance: 2.66 -> 2.67 | No clear effect |
| add optional consistency-loss training scheme | MVAdapter+CorrAdapter · Image-conditioned multi-view (GSO) | cPSNR: 19.68 -> 20.47 | Improved |
| add optional training scheme | MVAdapter+CorrAdapter · Text-conditioned multi-view (Objaverse) | FID: 23.42 -> 24.17; IS 15.96 -> 17.07 | Mixed |
| add CorrAdapter | Wan2.1-1.3B · Text-to-video (VBench) | Subject Consistency: 0.9536 -> 0.9715 | Improved |
| add CorrAdapter | Wan2.1-1.3B · Text-to-video (VBench) | Dynamic Degree: 0.5556 -> 0.5139 | Hurt |
| train CorrAdapter end-to-end with consistency loss instead of fine-tuning first then adding module (ablation) | MVAdapter+CorrAdapter · Image-conditioned multi-view (GSO) | PSNR: 24.05 -> 23.65 | Hurt |
| add CorrAdapter | MVAdapter · Runtime | peak memory (GB): 15.27 -> 20.86 | Hurt |
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.
- SyncDreamer — SyncDreamer: Generating Multiview-consistent Images from a Single-view Image (ICLR 2024) · better on 5 of 5 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
- Coupled Diffusion Sampling for Training-Free Multi-View Image Editing (CVPR 2026 · similarity 0.89)
- Color Alignment in Diffusion (CVPR 2025 · similarity 0.88)
- Inference-Time Text-to-Video Alignment with Diffusion Latent Beam Search (NeurIPS 2025 · similarity 0.88)
- SyncDreamer: Generating Multiview-consistent Images from a Single-view Image (ICLR 2024 · similarity 0.88)
- REPA Works Until It Doesn’t: Early-Stopped, Holistic Alignment Supercharges Diffusion Training (NeurIPS 2025 · similarity 0.88)
- Vivid-ZOO: Multi-View Video Generation with Diffusion Model (NeurIPS 2024 · similarity 0.87)
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.
