Research map / Video / Video generation
MoMaps: Semantics-Aware Scene Motion Generation with Motion Maps
Jiahui Lei, Kyle Genova, George Kopanas, Noah Snavely, Leonidas Guibas
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 |
|---|---|---|---|
| compress MoMaps with a fine-tuned VAE into latents instead of diffusing in raw MoMap space (ablation) | MoMap diffusion · BRIDGE | fg_mask_iou: 0.1305 -> 0.8128 | Improved |
| initialize U-Net from Stable Diffusion weights (ablation) | MoMap diffusion · BRIDGE | ate_dtw: 0.0792 -> 0.0689 | Improved |
| initialize both VAE and U-Net from pretrained weights (ablation) | MoMap diffusion · BRIDGE | ate_dtw: 0.0992 -> 0.0689 | Improved |
| add first-frame XYZ map condition (ablation) | MoMap diffusion · BRIDGE | ate_dtw: 0.0738 -> 0.0689 | Improved |
| add instance segmentation (semantic) image condition (ablation) | MoMap diffusion · BRIDGE | fg_mask_iou: 0.8079 -> 0.8128 | No clear effect |
| diffusion U-Net instead of conditional VAE for MoMap generation | MoMap generation · HOI4D | fg_mask_iou: 0.3400 -> 0.4492 | Improved |
| diffusion U-Net instead of conditional VAE on BRIDGE foreground mask | MoMap generation · BRIDGE | fg_mask_iou: 0.8271 -> 0.8128 | Hurt |
| MoMap diffusion vs GeneralFlow variants on direction accuracy | 3D motion generation · BRIDGE | quantize_acc_16: 0.7294 -> 0.7752 | Improved |
| MoMap diffusion vs GFlow-Rollout on local distance | HOI4D · HOI4D | local_dist_diff: 0.0086 vs 0.0092 | No clear effect |
Closest papers by meaning
- MoMask: Generative Masked Modeling of 3D Human Motions (CVPR 2024 · similarity 0.86)
- A 3D Scene is Worth 1K Tokens: 3D-Grounded Representation for Scene Generation at Scale (NeurIPS 2026 · similarity 0.86)
- HIMMEL: Hierarchical Interleaved Multi-stream Motion Encoding for Long Video Understanding (NeurIPS 2026 · similarity 0.86)
- MotionRAG: Motion Retrieval-Augmented Image-to-Video Generation (NeurIPS 2025 · similarity 0.86)
- Director3D: Real-world Camera Trajectory and 3D Scene Generation from Text (NeurIPS 2024 · similarity 0.86)
- VideoJAM: Joint Appearance-Motion Representations for Enhanced Motion Generation in Video Models (ICML 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.
