Research map / Detection and segmentation / Tracking
Multi-Object System Identification from Videos
Chunjiang Liu, Xiaoyuan Wang, Qingran Lin, Albert Xiao, Haoyu Chen, Shizheng Wen, Hao Zhang, Lu Qi, Ming-Hsuan Yang, Laszlo A. Jeni, Min Xu, Yizhou 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 |
|---|---|---|---|
| optimize continuous per-object material parameters instead of selecting from a fixed material library | MOSIV vs OmniPhysGS-RGB · MOSIV synthetic dataset, observable simulation | PSNR: 25.93 -> 30.51 | Improved |
| optimize continuous per-object parameters | MOSIV vs OmniPhysGS-RGB · MOSIV synthetic dataset, future simulation | PSNR: 19.00 -> 28.26 | Improved |
| give OmniPhysGS-RGB the ground-truth per-object material models (oracle) (ablation) | OmniPhysGS-RGB · MOSIV synthetic dataset, observable simulation | Chamfer distance: 11.79 -> 43.50 | Hurt |
| use object-wise Chamfer and silhouette losses instead of scene-wise losses (ablation) | MOSIV · MOSIV 6-scene subset | Chamfer distance: 22.13 -> 0.696 | Improved |
| use object-wise Chamfer and silhouette losses instead of scene-wise losses (ablation) | MOSIV · MOSIV 6-scene subset | PSNR: 27.89 -> 30.24 | Improved |
| add silhouette alpha loss to object-aware Chamfer loss (ablation) | MOSIV · MOSIV 6-scene subset | Chamfer distance: 1.17 -> 0.696 | Improved |
| add Chamfer loss to object-aware silhouette loss (ablation) | MOSIV · MOSIV 6-scene subset | Chamfer distance: 0.985 -> 0.696 | Improved |
| use MOSIV rather than reproduced CoupNeRF implicit representation | MOSIV vs CoupNeRF* · MOSIV dataset, future simulation | PSNR: 20.62 -> 28.92 | Improved |
| use MOSIV with Gaussian-based explicit representation instead of CoupNeRF | MOSIV · training cost per 30-frame sequence | Training time (s): 9591.09 -> 5021.46 | Improved |
| increase noise added to reconstructed Gaussians from 0.005 to 0.05 (ablation) | MOSIV · sensitivity, i.i.d. noise, observable simulation | PSNR: 27.86 -> 23.73 | Hurt |
| extend to three-object scenes | MOSIV · MOSIV dataset extension (3 objects) | PSNR: observable 23.98, future 19.81 | Hurt |
Closest papers by meaning
- Toward Material-Agnostic System Identification from Videos (ICCV 2025 · similarity 0.87)
- gradSim: Differentiable simulation for system identification and visuomotor control (ICLR 2021 · similarity 0.84)
- M^3-VOS: Multi-Phase, Multi-Transition, and Multi-Scenery Video Object Segmentation (CVPR 2025 · similarity 0.84)
- Vid2Sim: Generalizable, Video-based Reconstruction of Appearance, Geometry and Physics for Mesh-free Simulation (CVPR 2025 · similarity 0.84)
- Multiple Object Tracking as ID Prediction (CVPR 2025 · similarity 0.83)
- MOVIS: Enhancing Multi-Object Novel View Synthesis for Indoor Scenes (CVPR 2025 · similarity 0.82)
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.
