Research map / 3D vision / Human pose estimation
CUPID: Generative 3D Reconstruction via Joint Object and Pose Modeling
Binbin Huang, Haobin Duan, Yiqun Zhao, Zibo Zhao, Yi Ma, Shenghua Gao
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 low-level convolutional visual features to pixel-aligned DINOv2 conditioning (ablation) | Cupid refinement stage · Toys4K (sampled geometry and pose) | PSNR: 27.74 -> 30.05 | Improved |
| add pose-aligned conditioning (DINOv2 latent volume) to TRELLIS baseline (ablation) | Cupid refinement stage · Toys4K (GT geometry and pose) | PSNR: 31.84 -> 32.37 | Improved |
| add DINOv2 positional embeddings to SLat latents (ablation) | Cupid refinement stage · Toys4K (sampled geometry and pose) | PSNR: 27.47 -> 27.56 | No clear effect |
| zero latents of occluded voxels in the input view (ablation) | Cupid refinement stage · Toys4K (GT geometry and pose) | PSNR: 32.37 -> 32.39 | No clear effect |
| zero latents of occluded voxels in the input view (ablation) | Cupid refinement stage · Toys4K (sampled geometry and pose) | PSNR: 27.85 -> 27.74 | No clear effect |
| use full pose-aligned conditioning with sampled stage-one outputs (ablation) | Cupid · Toys4K | PSNR: 27.47 -> 30.05 | Improved |
| jointly generate object and camera pose, then inject pixel-aligned features, instead of OnePoseGen post-hoc alignment | Cupid · Toys4K input-view consistency | PSNR: 17.43 -> 30.05 | Improved |
| jointly generate object and camera pose instead of view-centric reconstruction (OpenLRM) | Cupid · GSO monocular geometry | Chamfer Distance (avg): 3.741 -> 1.823 | Improved |
| compare Cupid to point-map regression methods that use only visible geometry | Cupid · Toys4K monocular geometry | Chamfer Distance (avg): 2.534 vs 1.144 (VGGT), 1.284 (MoGe) | Hurt |
| use Cupid pixel-aligned conditioning instead of TRELLIS global conditioning | TRELLIS · Toys4K novel-view CLIP score (ViT-B/16) | CLIP image score: 0.9465 -> 0.9501 | Improved |
| add random occlusion masks on conditioning images during fine-tuning | Cupid · multi-object scenes | scene reconstruction quality | Improved |
Later papers that compared with this method
1 later accepted paper reports Cupid as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Global Pose Control for Generative View Synthesis in Normalized Object Coordinate Space (ECCV 2026 · better on 2 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- MoGe — MoGe: Unlocking Accurate Monocular Geometry Estimation for Open-Domain Images with Optimal Training Supervision (CVPR 2025) · better on 0 of 2 shared results
- LaRa — LaRa: Efficient Large-Baseline Radiance Fields (ECCV 2024) · 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
- Sparse-view Pose Estimation and Reconstruction via Analysis by Generative Synthesis (NeurIPS 2024 · similarity 0.87)
- Ouroboros3D: Image-to-3D Generation via 3D-aware Recursive Diffusion (CVPR 2025 · similarity 0.86)
- Decompositional Neural Scene Reconstruction with Generative Diffusion Prior (CVPR 2025 · similarity 0.86)
- SyncHuman: Synchronizing 2D and 3D Generative Models for Single-view Human Reconstruction (NeurIPS 2025 · similarity 0.86)
- GenRec: Knowing Where to Reconstruct and Where to Generate (NeurIPS 2026 · similarity 0.86)
- MVD-Fusion: Single-view 3D via Depth-consistent Multi-view Generation (CVPR 2024 · similarity 0.85)
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.
