Research map / 3D vision / 3D generation
Text-to-3D by Stitching a Multi-view Reconstruction Network to a Video Generator
Hyojun Go, Dominik Narnhofer, Goutam Bhat, Prune Truong, Federico Tombari, Konrad Schindler
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 model stitching of a 3D decoder to video VAE (Wan + AnySplat vs AnySplat alone) (ablation) | AnySplat · RealEstate10K novel view synthesis | PSNR: 20.85 -> 21.29 | Improved |
| add multi-view finetuning to the pretrained video model (ablation) | Wan + AnySplat · SceneBench | Imaging quality: 50.56 -> 54.56 | Improved |
| add 3D-consistency reward only to multi-view finetuning (ablation) | Wan + AnySplat · SceneBench | Imaging quality: 54.56 -> 38.67 | Hurt |
| add quality reward to multi-view finetuning (ablation) | Wan + AnySplat · SceneBench | Imaging quality: 54.56 -> 62.27 | Improved |
| add both quality and consistency rewards (full direct reward finetuning) over quality reward only (ablation) | Wan + AnySplat · SceneBench | Imaging quality: 62.27 -> 64.87 | Improved |
| add both rewards over quality reward only (ablation) | Wan + AnySplat · SceneBench | Aesthetic: 58.23 -> 56.96 | Hurt |
| add direct reward finetuning over multi-view only (ablation) | Wan + AnySplat · SceneBench | Unified Reward Alignment: 3.622 -> 3.667 | Improved |
| choose stitching layer with lower linear stitching MSE (ablation) | Wan + VGGT · 7-Scenes pointmap | Acc./Comp./NC | Improved |
| use CKA instead of MSE to pick the stitching layer (ablation) | Wan + VGGT · 7-Scenes pointmap | best layer identification | Hurt |
| use stitched latent decoder instead of sequential decode-then-VGGT under latent noise (ablation) | stitched VGGT · ETH3D pointmap | Acc./Comp./NC under noise | Improved |
| stitch Wan VAE to VGGT instead of using VGGT on images | VGGT · 7-Scenes pointmap | Accuracy mean: 0.020 -> 0.018 | No clear effect |
| stitch Wan VAE to MVDUSt3R instead of using MVDUSt3R on images | MVDUSt3R · ETH3D pointmap | Accuracy mean: 0.400 -> 0.401 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- VideoRFSplat — VideoRFSplat: Direct Scene-Level Text-to-3D Gaussian Splatting Generation with Flexible Pose and Multi-View Joint Modeling (ICCV 2025) · better on 5 of 5 shared results
- Director3D — Director3D: Real-world Camera Trajectory and 3D Scene Generation from Text (NeurIPS 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
- Let 2D Diffusion Model Know 3D-Consistency for Robust Text-to-3D Generation (ICLR 2024 · similarity 0.88)
- IT3D: Improved Text-to-3D Generation with Explicit View Synthesis (AAAI 2024 · similarity 0.87)
- DreamDPO: Aligning Text-to-3D Generation with Human Preferences via Direct Preference Optimization (ICML 2025 · similarity 0.87)
- MARVEL-40M+: Multi-Level Visual Elaboration for High-Fidelity Text-to-3D Content Creation (CVPR 2025 · similarity 0.87)
- IM-3D: Iterative Multiview Diffusion and Reconstruction for High-Quality 3D Generation (ICML 2024 · similarity 0.87)
- Instant3D: Fast Text-to-3D with Sparse-view Generation and Large Reconstruction Model (ICLR 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.
