Research map / 3D vision / Depth and stereo
The Dynamic Prior: Understanding 3D Structures for Casual Dynamic Videos
Zhuoyuan Wu, Xurui Yang, Jiahui Huang, Yue Wang, Jun Gao
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 |
|---|---|---|---|
| use GPT-4o instead of Qwen2.5-VL-7B for dynamic object reasoning (ablation) | Dynapo · FBMS-59 | J&F: 82.27 -> 86.47 | Improved |
| scale reasoner from Qwen2.5-VL-7B to 32B (ablation) | Dynapo · FBMS-59 | J&F: 82.27 -> 83.80 | Improved |
| scale reasoner from InternVL3.5-8B to 38B (ablation) | Dynapo · FBMS-59 | J&F: 78.36 -> 83.31 | Improved |
| use 16 keyframes instead of 32 (ablation) | Dynapo · FBMS-59 | J&F: 82.35 -> 86.47 | Improved |
| scale segmentation MLLM from InternVL2.5-1B to 4B (ablation) | Dynapo · FBMS-59 | J&F: 83.40 -> 87.71 | Improved |
| scale segmentation MLLM from InternVL2.5-4B to 26B (ablation) | Dynapo · FBMS-59 | J&F: 87.71 -> 83.04 | Hurt |
| add Dynapo mask in bundle adjustment (vs no mask) (ablation) | MegaSam · LightSpeed pose | ATE: 0.090 -> 0.052 | Improved |
| use Dynapo mask instead of method's own mask in BA (ablation) | MegaSam · LightSpeed pose | ATE: 0.072 -> 0.052 | Improved |
| use Dynapo mask instead of MegaSam mask in BA (ablation) | MegaSam · LightSpeed pose | RRE: 2.504 -> 0.910 | Improved |
| add BA without any mask (ablation) | MegaSam · LightSpeed pose | ATE: 0.073 -> 0.090 | Hurt |
| add Dynapo mask in BA (vs no mask) (ablation) | VGGT · LightSpeed pose | ATE: 0.182 -> 0.178 | No clear effect |
| add Dynapo mask in BA (vs no mask) (ablation) | SpatialTrackerv2 · LightSpeed pose | ATE: 0.168 -> 0.103 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MegaSam — MegaSaM: Accurate, Fast and Robust Structure and Motion from Casual Dynamic Videos (CVPR 2025) · 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
- 4D3R: Motion-Aware Neural Reconstruction and Rendering of Dynamic Scenes from Monocular Videos (NeurIPS 2025 · similarity 0.90)
- OSN: Infinite Representations of Dynamic 3D Scenes from Monocular Videos (ICML 2024 · similarity 0.89)
- MegaSaM: Accurate, Fast and Robust Structure and Motion from Casual Dynamic Videos (CVPR 2025 · similarity 0.89)
- MonST3R: A Simple Approach for Estimating Geometry in the Presence of Motion (ICLR 2025 · similarity 0.88)
- S3O: A Dual-Phase Approach for Reconstructing Dynamic Shape and Skeleton of Articulated Objects from Single Monocular Video (ICML 2024 · similarity 0.88)
- Stereo4D: Learning How Things Move in 3D from Internet Stereo Videos (CVPR 2025 · similarity 0.88)
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.
