Research map / Video / Video generation
GeCo: Evaluating Geometric Consistency for Video Generation via Motion and Structure
Leslie Gu, Junhwa Hur, Charles Herrmann, Fangneng Zhan, Todd Zickler, Deqing Sun, Hanspeter Pfister
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 structure cue to motion cue (fusion) (ablation) | GeCo · OccluBench inconsistency segmentation | AP (%): 4.43 -> 83.48 | Improved |
| add structure cue to motion cue (fusion) (ablation) | GeCo · WarpBench CO3D-Warp localization | AP (%): 64.90 -> 60.63 | Hurt |
| add structure cue to motion cue (fusion) (ablation) | GeCo · WarpBench ScanNet-Warp anomaly detection | accuracy (%): 89.23 -> 92.31 | Improved |
| add structure cue to motion cue (fusion) (ablation) | GeCo · WarpBench CO3D-Warp anomaly detection | accuracy (%): 71.59 -> 55.68 | Hurt |
| add motion cue to structure cue (fusion) (ablation) | GeCo · WarpBench CO3D-Warp localization | AP (%): 25.64 -> 60.63 | Improved |
| add GeCo guidance to CogVideoX-5B sampling | CogVideoX-5B · 45 prompts | GeCo Fused score: 0.147 -> 0.135 | Improved |
| add GeCo guidance to CogVideoX-5B sampling | CogVideoX-5B · 45 prompts | MEt3R: 0.100 -> 0.098 | Improved |
| add GeCo guidance to CogVideoX-5B sampling | CogVideoX-5B · 45 prompts | Total Motion (%): 74.35 -> 83.16 | Improved |
| use predicted instead of ground-truth geometry | GeCo · TartanAir V2 Hard | Fused score: 0.0024 -> 0.0109 | No clear effect |
| use MEt3R instead of GeCo on real vs generated | MEt3R · DL3DV real vs generated | Cohen's d: -0.361 | Hurt |
Closest papers by meaning
- Measuring 3D Spatial Geometric Consistency in Dynamic Video Generation (ECCV 2026 · similarity 0.92)
- Learning to Correct Geometry in Generated Videos (NeurIPS 2026 · similarity 0.89)
- SimVS: Simulating World Inconsistencies for Robust View Synthesis (CVPR 2025 · similarity 0.88)
- PanoWorld: Geometry-Consistent Panoramic Video World Modeling (NeurIPS 2026 · similarity 0.87)
- GeoVideo: Introducing Geometric Regularization into Video Generation Model (NeurIPS 2025 · similarity 0.87)
- GEN3C: 3D-Informed World-Consistent Video Generation with Precise Camera Control (CVPR 2025 · 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.
