Research map / Video / Video segmentation and tracking
Visual Interaction Networks: Learning a Physics Simulator from Video
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 relation network to IN dynamics cores (ablation) | VIN · 5 physical systems (3 and 6 objects) | Inverse Normalized Loss | Helped |
| use Interaction Network dynamics predictor instead of MLP (Visual RNN) or LSTM (Visual LSTM) | visual encoder + dynamics predictor · 5 physical systems | Inverse Normalized Loss | Helped |
| learn dynamics from pixels (noisy encoded states) instead of ground-truth states | Interaction Network / LSTM · 3-object datasets, 50-step rollouts | Euclidean prediction error | Helped |
Closest papers by meaning
- Learning Long-term Visual Dynamics with Region Proposal Interaction Networks (ICLR 2021 · similarity 0.91)
- Visual Grounding of Learned Physical Models (ICML 2020 · similarity 0.90)
- InterDyn: Controllable Interactive Dynamics with Video Diffusion Models (CVPR 2025 · similarity 0.90)
- Unsupervised Learning for Physical Interaction through Video Prediction (NeurIPS 2016 · similarity 0.90)
- Physion: Evaluating Physical Prediction from Vision in Humans and Machines (NeurIPS 2021 · similarity 0.89)
- Interaction Networks for Learning about Objects, Relations and Physics (NeurIPS 2016 · similarity 0.89)
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.