Research map / 3D vision / 3D generation
Aligned Novel View Image and Geometry Synthesis via Cross-modal Attention Instillation
Min-Seop Kwak, Junho Kim, Sangdoo Yun, Dongyoon Han, Taekyung Kim, Seungryong Kim, Jin-Hwa Kim
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 pointmap conditioning to baseline (ablation) | image diffusion NVS · RealEstate10K extrapolative | PSNR: 16.55 -> 16.93 | Improved |
| add proximity-based mesh conditioning (ablation) | pointmap-conditioned model · RealEstate10K extrapolative | PSNR: 16.93 -> 17.01 | Improved |
| add cross-modal attention instillation (ablation) | full model · RealEstate10K extrapolative | PSNR: 17.01 -> 17.41 | Improved |
| add all components to baseline (ablation) | image diffusion NVS · RealEstate10K extrapolative | LPIPS: 0.260 -> 0.229 | Improved |
| increase reference views from 2 to 3 at inference (trained on 2) | Ours · RealEstate10K | PSNR: 17.41 -> 20.02 | Improved |
| increase reference views from 3 to 4 | Ours · RealEstate10K | PSNR: 20.02 -> 20.08 | No clear effect |
| increase reference views from 1 to 7 | Ours · Co3D | PSNR: 14.62 -> 23.45 | Improved |
| add Gaussian perturbation sigma 15% to pointmap | Ours · DTU extrapolative 2-view | PSNR: 15.580 -> 13.726 | Hurt |
| mask 80% of predicted pointmap points | Ours · DTU extrapolative 2-view | PSNR: 15.580 -> 13.000 | Hurt |
| mask 30% of predicted pointmap points | Ours · DTU extrapolative 2-view | LPIPS: 0.184 -> 0.223 | Hurt |
| use Ours instead of NoPoSplat in interpolative setting | Ours vs NoPoSplat · RealEstate10K interpolative | PSNR: 25.03 -> 24.23 | Hurt |
| use Ours instead of LVSM (2-view) | Ours · DTU extrapolative | PSNR: 15.23 -> 15.58 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 4 of 4 on most shared results.
- PixelSplat — pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstruction (CVPR 2024) · better on 6 of 6 shared results
- MVSplat — MVSplat: Efficient 3D Gaussian Splatting from Sparse Multi-View Images (ECCV 2024) · better on 6 of 6 shared results
- LucidDreamer — LucidDreamer: Towards High-Fidelity Text-to-3D Generation via Interval Score Matching (CVPR 2024) · better on 3 of 3 shared results
- GenWarp — GenWarp: Single Image to Novel Views with Semantic-Preserving Generative Warping (NeurIPS 2024) · better on 3 of 3 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
- Novel View Synthesis with Pixel-Space Diffusion Models (CVPR 2025 · similarity 0.89)
- Enhancing Novel View Synthesis via Geometry Grounded Set Diffusion (NeurIPS 2026 · similarity 0.89)
- Efficient-3Dim: Learning a Generalizable Single-image Novel-view Synthesizer in One Day (ICLR 2024 · similarity 0.88)
- Era3D: High-Resolution Multiview Diffusion using Efficient Row-wise Attention (NeurIPS 2024 · similarity 0.88)
- Zero-Shot Novel View and Depth Synthesis with Multi-View Geometric Diffusion (CVPR 2025 · similarity 0.88)
- NAMVIS: Next-Scale Autoregressive Multi-View Image Synthesis (NeurIPS 2026 · similarity 0.88)
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.
