Research map / 3D vision / 3D generation
GenMatter: Perceiving Physical Objects with Generative Matter Models
Eric Li, Arijit Dasgupta, Yoni Friedman, Mathieu Huot, Vikash Mansinghka, Thomas O'Connell, William T. Freeman, Joshua B. Tenenbaum
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 cluster layer (hierarchy) over particles, versus fixed particles K=5 (ablation) | GenMatter · Random dot kinematograms | correlation r^2: 0.35 -> 0.86 | Improved |
| add cluster layer over adaptive particles K=500 (ablation) | GenMatter · Random dot kinematograms | correlation r^2: 0.41 -> 0.86 | Improved |
| add cluster-level variables (ablation) | GenMatter · TAP-Vid DAVIS (SAM init) | matter-weighted Jaccard: 0.69 -> 0.79 | Improved |
| add monocular depth input (ablation) | GenMatter · TAP-Vid DAVIS (SAM init) | matter-weighted Jaccard: 0.69 -> 0.79 | Improved |
| add monocular depth input (ablation) | GenMatter · Gestalt camouflaged videos | per-pixel accuracy: 0.89 -> 0.94 | Improved |
| use ground-truth mask instead of SAM2 for initialization | GenMatter · TAP-Vid DAVIS | matter-weighted Jaccard: 0.79 -> 0.77 | Hurt |
| use ground-truth mask instead of SAM for initialization | CoTracker3 · TAP-Vid DAVIS | matter-weighted Jaccard: 0.78 -> 0.78 | No clear effect |
| replace FlowSAM with GenMatter | FlowSAM · Gestalt camouflaged videos | Jaccard: 0.67 -> 0.72 | Improved |
| replace SegAnyMo with GenMatter | SegAnyMo · Gestalt camouflaged videos | Jaccard: 0.26 -> 0.72 | Improved |
| subsample data points to 1/128 of full resolution | GenMatter · TAP-Vid DAVIS | matter-weighted Jaccard: 0.76 -> 0.79 | No clear effect |
| subsample data points to 1/512 of full resolution | GenMatter · TAP-Vid DAVIS | matter-weighted Jaccard: 0.76 -> 0.56 | Hurt |
| subsample data points from full resolution to 1/128 | GenMatter · TAP-Vid DAVIS | FPS: 0.80 -> 9.8 | Improved |
Closest papers by meaning
- Generative Perception of Shape and Material from Differential Motion (NeurIPS 2025 · similarity 0.90)
- Dynamics: Language-Based Representation for Inferring Rigid-Body Dynamics From Videos (CVPR 2026 · similarity 0.88)
- Object segmentation from common fate: Motion energy processing enables human-like zero-shot generalization to random dot stimuli (NeurIPS 2024 · similarity 0.87)
- Compositional Scene Understanding through Inverse Generative Modeling (ICML 2025 · similarity 0.86)
- PhysCtrl: Generative Physics for Controllable and Physics-Grounded Video Generation (NeurIPS 2025 · 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.
