Research map / 3D vision / 3D generation
DiST-4D: Disentangled Spatiotemporal Diffusion with Metric Depth for 4D Driving Scene Generation
Jiazhe Guo, Yikang Ding, Xiwu Chen, Shuo Chen, Bohan Li, Yingshuang Zou, Xiaoyang Lyu, Feiyang Tan, Xiaojuan Qi, Zhiheng Li, Hao Zhao
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 self-supervised cycle consistency (SCC) fine-tuning (ablation) | DiST-S · nuScenes NVS, shift 1m | FID: 12.00 -> 10.12 | Improved |
| add SCC fine-tuning (ablation) | DiST-S · nuScenes NVS, shift 4m | FVD: 137.03 -> 105.29 | Improved |
| add SCC fine-tuning (ablation) | DiST-4D · nuScenes NVS, shift 4m | FID: 33.56 -> 25.16 | Improved |
| condition on generated DiST-T videos instead of real videos with metric depth | DiST-S · nuScenes NVS, shift 1m | FID: 12.00 -> 20.64 | Hurt |
| add valid mask to depth conditioning (ablation) | DiST-S · DiST-S ablation FID-1m | FID: 30.31 -> 25.51 | Improved |
| add depth modality without valid mask (ablation) | DiST-S · DiST-S ablation FID-1m | FID: 26.33 -> 30.31 | Hurt |
| add data augmentation (n=+-4 projection and downsampling) (ablation) | DiST-S · DiST-S ablation FID-1m | FID: 26.19 -> 25.51 | Improved |
| add reference depth as input to DiST-T (ablation) | DiST-T · nuScenes val video generation | FID: 6.83 -> 7.40 | Hurt |
| add reference depth as input to DiST-T (ablation) | DiST-T · nuScenes depth generation, multi-frame GT | Abs. Rel.: 0.24 -> 0.20 | Improved |
| use DiST-T instead of UniScene for video generation | UniScene · nuScenes val video generation | FVD: 71.94 -> 22.67 | Improved |
| use DiST-S (+SCC) instead of StreetGaussian | StreetGaussian · nuScenes NVS, shift 1m | FID: 32.12 -> 10.12 | 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.
- FreeVS* — FreeVS: Generative View Synthesis on Free Driving Trajectory (ICLR 2025) · better on 1 of 1 shared result
- M2Depth — M^2Depth: Self-supervised Two-Frame Multi-camera Metric Depth Estimation (ECCV 2024) · better on 1 of 1 shared result
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
- Diffusion4D: Fast Spatial-temporal Consistent 4D generation via Video Diffusion Models (NeurIPS 2024 · similarity 0.91)
- 4Diffusion: Multi-view Video Diffusion Model for 4D Generation (NeurIPS 2024 · similarity 0.90)
- DreamScene4D: Dynamic Multi-Object Scene Generation from Monocular Videos (NeurIPS 2024 · similarity 0.88)
- Controlling Space and Time with Diffusion Models (ICLR 2025 · similarity 0.87)
- A 3D Scene is Worth 1K Tokens: 3D-Grounded Representation for Scene Generation at Scale (NeurIPS 2026 · similarity 0.87)
- CAT4D: Create Anything in 4D with Multi-View Video Diffusion Models (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.
