Research map / Video / Video generation
ConsistentCity: Semantic Flow-guided Occupancy DiT for Temporally Consistent Driving Scene Synthesis
Benjin Zhu, Xiaogang Wang, Hongsheng Li
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 semantic flow-guided feature warping (ablation) | SF-DiT baseline · nuScenes video synthesis | FID: 15.4 -> 11.2 | Improved |
| add semantic flow-guided feature warping (ablation) | SF-DiT baseline · nuScenes video synthesis | FVD: 94.2 -> 81.3 | Improved |
| add adaptive attention modulation (ablation) | SF-DiT with feature warping · nuScenes video synthesis | FID: 11.2 -> 9.4 | Improved |
| add FlowResEstimator (ablation) | SF-DiT with attention modulation · nuScenes video synthesis | FID: 9.4 -> 8.9 | Improved |
| add flow uncertainty (ablation) | SF-DiT with FlowResEstimator · nuScenes video synthesis | FID: 8.9 -> 8.5 | Improved |
| add flow classification (ablation) | SF-DiT with uncertainty · nuScenes video synthesis | FID: 8.5 -> 8.3 | Improved |
| use SFMCA instead of spatial-temporal self-attention (ablation) | SF-DiT · nuScenes video synthesis | FVD: 235.1 -> 73.6 | Improved |
| use SFMCA instead of spatial self + temporal cross attention (ablation) | SF-DiT · nuScenes video synthesis | FVD: 94.2 -> 73.6 | Improved |
| add depth to semantics conditioning (ablation) | ControlNet with SFMCA · nuScenes video synthesis | FID: 12.8 -> 9.1 | Improved |
| add flow to depth conditioning (ablation) | ControlNet with SFMCA · nuScenes video synthesis | FID: 13.9 -> 11.4 | Improved |
| add flow to semantics+depth conditioning (ablation) | ControlNet with SFMCA · nuScenes video synthesis | FID: 9.1 -> 8.3 | Improved |
| increase sequence length from 4 to 8 frames (ablation) | SF-DiT · nuScenes video synthesis | FID: 10.2 -> 8.3 | Improved |
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.
- Panacea — Panacea: Panoramic and Controllable Video Generation for Autonomous Driving (CVPR 2024) · better on 2 of 2 shared results
- DriveDreamer — DriveDreamer: Towards Real-world-driven World Models for Autonomous Driving (ECCV 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
- X-Scene: Large-Scale Driving Scene Generation with High Fidelity and Flexible Controllability (NeurIPS 2025 · similarity 0.88)
- Learning Temporal 3D Semantic Scene Completion via Optical Flow Guidance (NeurIPS 2025 · similarity 0.88)
- FlowVid: Taming Imperfect Optical Flows for Consistent Video-to-Video Synthesis (CVPR 2024 · similarity 0.87)
- UniScene: Unified Occupancy-centric Driving Scene Generation (CVPR 2025 · similarity 0.87)
- Genesis: Multimodal Driving Scene Generation with Spatio-Temporal and Cross-Modal Consistency (NeurIPS 2025 · similarity 0.87)
- Achieving Speed-Accuracy Balance in Vision-based 3D Occupancy Prediction via Geometric-Semantic Disentanglement (AAAI 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.
