Research map / Video / Video generation
Measuring 3D Spatial Geometric Consistency in Dynamic Video Generation
Weijia Dou, Wenzhao Zheng, Weiliang Chen, Yu Zheng, Jie Zhou, Jiwen Lu
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 moving object segmentation (ablation) | SGC · RT-1 | SGC: 0.163 -> 0.064 | Improved |
| add moving object segmentation (ablation) | SGC · OpenVid | SGC: 0.014 -> 0.053 | Hurt |
| add depth-aware plane clustering instead of spatial grid (ablation) | SGC · nuScenes | SGC: 0.163 -> 0.061 | Improved |
| use 1D depth clustering instead of depth plus spatial contiguity (ablation) | SGC · nuScenes | SGC: 0.153 -> 0.061 | Improved |
| use 1D depth clustering instead of depth plus spatial contiguity (ablation) | SGC · Latte | SGC: 0.555 -> 0.323 | Improved |
| use VGGT instead of dynamic-scene estimators Any4D or Page4D (ablation) | SGC · Estimator robustness | Spearman rho vs original: 0.860 | No clear effect |
| increase warping perturbation severity | SGC · nuScenes | SGC: 0.0613 -> 0.0979 | Hurt |
| increase perspective squeeze severity | SGC · nuScenes | SGC: 0.0613 -> 0.2401 | Hurt |
| increase spatial depth warp severity | SGC · nuScenes | SGC: 0.0613 -> 0.2252 | Hurt |
| apply static background mask to MEt3R | MEt3R · generated videos | MEt3R error: consistent upward shift in error scores | Hurt |
| use SGC instead of MEt3R(+MOS) | metric choice · human judgments | Spearman rho: 0.773 -> 0.945 | Improved |
| use SGC instead of VBench-BC | metric choice · human judgments | Spearman rho: 0.400 -> 0.945 | Improved |
Closest papers by meaning
- GeCo: Evaluating Geometric Consistency for Video Generation via Motion and Structure (ECCV 2026 · similarity 0.92)
- Learning to Correct Geometry in Generated Videos (NeurIPS 2026 · similarity 0.90)
- PanoWorld: Geometry-Consistent Panoramic Video World Modeling (NeurIPS 2026 · similarity 0.87)
- SimVS: Simulating World Inconsistencies for Robust View Synthesis (CVPR 2025 · similarity 0.87)
- GeoVideo: Introducing Geometric Regularization into Video Generation Model (NeurIPS 2025 · similarity 0.87)
- Direct Motion Models for Assessing Generated Videos (ICML 2025 · similarity 0.86)
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.
