Research map / 3D vision / 3D generation
Global Pose Control for Generative View Synthesis in Normalized Object Coordinate Space
Zhibing Li, Amogh Gupta, Behnoosh Parsa, Dan Casas
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 LoRA fine-tuning to off-the-shelf Qwen-Image-Edit with text camera prompts (ablation) | Qwen-Image-Edit · Camera conditioning ablation | PSNR: 12.48 -> 16.78 | Improved |
| Add Plucker camera tokens with global attention (ablation) | Baseline + LoRA · Camera conditioning ablation | PSNR: 16.78 -> 16.87 | No clear effect |
| Replace global attention with regional attention for camera tokens (ablation) | Baseline + LoRA + camera tokens · Camera conditioning ablation | PSNR: 16.87 -> 20.80 | Improved |
| Replace global attention with regional attention for camera tokens (ablation) | Baseline + LoRA + camera tokens · Camera conditioning ablation | LPIPS: 0.180 -> 0.111 | Improved |
| Add front face text definition to training (ablation) | Model trained on ABO only · unseen categories (qualitative) | qualitative | Improved |
| Add canonical front view anchoring among generated targets (ablation) | Ours · swords and thin objects (qualitative) | pose accuracy | Improved |
| Replace detailed captions with category-only captions (ablation) | Ours · Toys4k 1 view | PSNR: 18.24 -> 18.18 | No clear effect |
| Replace detailed captions with category-only captions (ablation) | Ours · Toys4k 5 views | PSNR: 19.68 -> 19.70 | No clear effect |
| Use global NOCS pose control instead of estimated relative input poses | Ours vs SEVA · GSO 1 view | PSNR: 15.20 -> 17.25 | Improved |
| Use GT input poses for SEVA instead of estimated poses (3 views) | SEVA vs Ours · Toys4k 3 views | PSNR: 21.35 -> 19.31 | Hurt |
| Add more input views from 1 to 5 | Ours · GSO | PSNR: 17.25 -> 19.07 | Improved |
| Use 2D image-editing NVS instead of image-to-3D generation | Ours vs Orientation Matters · Toys4k 1 view | PSNR: 16.94 -> 18.24 | Improved |
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.
- Cupid — CUPID: Generative 3D Reconstruction via Joint Object and Pose Modeling (CVPR 2026) · better on 2 of 2 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
- 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)
- NViST: In the Wild New View Synthesis from a Single Image with Transformers (CVPR 2024 · similarity 0.88)
- One View Is Enough: In-the-Wild Monocular Pretraining for Novel View Generation (NeurIPS 2026 · similarity 0.88)
- Free3D: Consistent Novel View Synthesis without 3D Representation (CVPR 2024 · similarity 0.87)
- OBJECT-UNI: A Unified Model for Object-Centric Spatial Understanding and Controllable Generation (NeurIPS 2026 · 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.
