Research map / 3D vision / 3D generation
WorldStereo: Bridging Camera-Guided Video Generation and Scene Reconstruction via 3D Geometric Memories
Yisu Zhang, Chenjie Cao, Tengfei Wang, Xuhui Zuo, Junta Wu, Jianke Zhu, Chunchao Guo
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 Global-Geometric Memory (ablation) | WorldStereo* · Tanks-and-Temples | F1-Score: 0.447 -> 0.485 | Improved |
| add Spatial-Stereo Memory on top of GGM (ablation) | WorldStereo-GGM · Tanks-and-Temples | F1-Score: 0.485 -> 0.578 | Improved |
| add Global-Geometric Memory (ablation) | WorldStereo* · MipNeRF360 | F1-Score: 0.350 -> 0.342 | No clear effect |
| add Spatial-Stereo Memory on top of GGM (ablation) | WorldStereo-GGM · MipNeRF360 | F1-Score: 0.342 -> 0.406 | Improved |
| add Global-Geometric Memory (ablation) | WorldStereo* · Memory-component benchmark | PSNR: 14.64 -> 17.45 | Improved |
| add Global-Geometric Memory (ablation) | WorldStereo* · Memory-component benchmark | RotErr: 1.300 -> 0.699 | Improved |
| add Spatial-Stereo Memory on top of GGM (ablation) | WorldStereo-GGM · Memory-component benchmark | PSNR: 17.45 -> 18.40 | Improved |
| add Spatial-Stereo Memory on top of GGM (ablation) | WorldStereo-GGM · WorldScore OOD benchmark | RotErr: 0.129 -> 0.145 | Hurt |
| distill with DMD (4 steps) | WorldStereo-Full · WorldScore OOD benchmark | inference time (sec): 173 -> 9 | Improved |
| distill with DMD (4 steps) | WorldStereo-Full · Tanks-and-Temples | F1-Score: 0.578 -> 0.534 | Hurt |
| order panorama trajectories orbit first instead of panorama-only memory (ablation) | WorldStereo memory bank · 3D panorama generation | overlapping FoV score (all): 167.2 -> 177.6 | Improved |
| add pointmap 3D correspondence to SSM input (ablation) | Spatial-Stereo Memory · qualitative ablation | consistency with retrieved references | 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.
- Gen3C — GEN3C: 3D-Informed World-Consistent Video Generation with Precise Camera Control (CVPR 2025) · 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
- AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories (ECCV 2026 · similarity 0.91)
- Gen3R: 3D Scene Generation Meets Feed-Forward Reconstruction (CVPR 2026 · similarity 0.90)
- Video World Models with Long-term Spatial Memory (NeurIPS 2025 · similarity 0.90)
- GEN3C: 3D-Informed World-Consistent Video Generation with Precise Camera Control (CVPR 2025 · similarity 0.89)
- PanoWorld: Geometry-Consistent Panoramic Video World Modeling (NeurIPS 2026 · similarity 0.89)
- World from Motion: Generative Dynamic Gaussian Reconstruction from Monocular Video (NeurIPS 2026 · similarity 0.89)
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.
