Research map / Image generation / Text-to-image diffusion
NVS-Adapter: Plug-and-Play Novel View Synthesis from a Single Image
Yoonwoo Jeong, Jinwoo Lee, Chiheon Kim, Minsu Cho, Doyup Lee
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 |
|---|---|---|---|
| generate 4 target views jointly instead of 1 (ablation) | NVS-Adapter · Google Scanned Objects (16 views) | PSNR: 14.38 -> 19.01 | Improved |
| increase target views from 4 to 6 (ablation) | NVS-Adapter · Google Scanned Objects | PSNR: 19.01 -> 18.38 | Hurt |
| increase learnable tokens from c=1 to c=2 (ablation) | target views alignment · Google Scanned Objects | PSNR: 18.56 -> 18.51 | No clear effect |
| add global semantic conditioning with CLIP image embedding (ablation) | NVS-Adapter · Google Scanned Objects | PSNR: 18.27 -> 18.45 | Improved |
| use Plucker ray embeddings instead of camera extrinsics (ablation) | view-consistency cross-attention · Google Scanned Objects | PSNR: 18.02 -> 18.45 | Improved |
| use Plucker instead of ray origin/direction embeddings (ablation) | view-consistency cross-attention · Google Scanned Objects | PSNR: 18.41 -> 18.45 | No clear effect |
| add HED ControlNet to NVS-Adapter without extra training | NVS-Adapter (SD1.5) · Google Scanned Objects | PSNR: 17.87 -> 22.77 | Improved |
| combine multiple ControlNet conditions (ablation) | NVS-Adapter (SD1.5) · Google Scanned Objects | PSNR: 22.77 -> 19.91 | Hurt |
| fine-tune all T2I parameters instead of adapter only (ablation) | NVS-Adapter · Google Scanned Objects | PSNR: 18.87 -> 19.00 | No clear effect |
| full cross-attention instead of learnable tokens (ablation) | target views alignment · Google Scanned Objects | PSNR: 18.87 -> 18.93 | No clear effect |
| compare to Zero123++ on its fixed viewpoints | NVS-Adapter · GSO (Zero123++ viewpoints) | PSNR: 21.64 -> 18.27 | Hurt |
Closest papers by meaning
- NVComposer: Boosting Generative Novel View Synthesis with Multiple Sparse and Unposed Images (CVPR 2025 · similarity 0.89)
- Novel View Synthesis with Pixel-Space Diffusion Models (CVPR 2025 · similarity 0.88)
- NVS-Solver: Video Diffusion Model as Zero-Shot Novel View Synthesizer (ICLR 2025 · similarity 0.87)
- TOSS: High-quality Text-guided Novel View Synthesis from a Single Image (ICLR 2024 · similarity 0.87)
- LVSM: A Large View Synthesis Model with Minimal 3D Inductive Bias (ICLR 2025 · similarity 0.86)
- G-NeRF: Geometry-enhanced Novel View Synthesis from Single-View Images (CVPR 2024 · similarity 0.86)
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.
