Research map / Image generation / Text-to-image diffusion
VERTIGO: Visual Preference Optimization for Cinematic Camera Generation
Mengtian Li, Yuwei Lu, Feifei Li, Chenqi Gan, Zhifeng Xie, Xi WANG
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 |
|---|---|---|---|
| post-train GenDoP with visual preference DPO (VERTIGO) | GenDoP · Unity render (1000 trajectories) | Missing rate (MisR): 0.387 -> 0.008 | Improved |
| post-train GenDoP with visual preference DPO (VERTIGO) | GenDoP · LenScript test set (geometry) | FCD: 4.17 -> 4.22 | No clear effect |
| post-train GenDoP with visual preference DPO (VERTIGO) | GenDoP · LenScript test set (geometry) | CLaTr-Score: 67.98 -> 68.40 | Improved |
| post-train GenDoP with visual preference DPO (VERTIGO) | GenDoP · Wan 2.1 VACE video generation | VBench aesthetic: 0.722 -> 0.735 | Improved |
| use cyclic semantic scoring instead of tag-consistency scoring for preference pairs (ablation) | VERTIGO · Unity render (scoring ablation) | Missing rate (MisR): 0.327 -> 0.008 | Improved |
| use cyclic semantic scoring instead of interpolation-trained direct regression scoring (ablation) | VERTIGO · Unity render (scoring ablation) | Missing rate (MisR): 0.286 -> 0.008 | Improved |
| use cyclic semantic scoring instead of tag scoring (ablation) | VERTIGO · Unity render (scoring ablation) | VBench consistency: 0.902 -> 0.908 | Improved |
| use DPO beta 0.1 instead of 0.01 (ablation) | VERTIGO · Unity render (DPO beta ablation) | Missing rate (MisR): 0.097 -> 0.008 | Improved |
| use DPO beta 0.1 instead of 0.5 (ablation) | VERTIGO · Unity render (DPO beta ablation) | Missing rate (MisR): 0.132 -> 0.008 | Improved |
| use DPO beta 0.1 instead of 0.9 (ablation) | VERTIGO · Unity render (DPO beta ablation) | Missing rate (MisR): 0.414 -> 0.008 | Improved |
| use DPO beta 0.1 instead of 0.01 (ablation) | VERTIGO · Unity render (DPO beta ablation) | VBench aesthetic: 0.482 -> 0.518 | Improved |
| apply VERTIGO post-training on animated characters | GenDoP · Unity render (animated characters) | Missing rate (MisR): 0.387 -> 0.008 | Improved |
Closest papers by meaning
- CineTechBench: A Benchmark for Cinematographic Technique Understanding and Generation (NeurIPS 2025 · similarity 0.88)
- Stable Cinemetrics : Structured Taxonomy and Evaluation for Professional Video Generation (NeurIPS 2025 · similarity 0.87)
- Unveiling the Value of Motion for Cinematic Camera Trajectories (NeurIPS 2026 · similarity 0.86)
- Auteur: Language-Driven Cinematographic Framing for Human-Centric Video Generation (NeurIPS 2026 · similarity 0.86)
- ShotBench: Expert-Level Cinematic Understanding in Vision-Language Models (NeurIPS 2025 · similarity 0.85)
- CineOrchestra: Unified Entity-Centric Conditioning for Cinematic Video Generation (NeurIPS 2026 · 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.
