Research map / Video / Video segmentation and tracking
Schroedinger’s Cat: Probabilistic Representation and Prediction of Potential Scene Kinematics
Timy Phan, Jannik Wiese, Bjorn Ommer
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 localized per-element latents instead of a global token (ablation) | point-wise decoder · held-out validation set | EPE: 0.479 -> 0.011 | Improved |
| add localized per-element latents instead of entangled latents (ablation) | point-wise decoder · held-out validation set | EPE: 0.509 -> 0.011 | Improved |
| use full joint decoder instead of point-wise decoder with 2 goals (ablation) | GARFIELD · held-out validation set | EPE: 0.035 -> 0.023 | Improved |
| use full joint decoder instead of point-wise decoder with 16 goals (ablation) | GARFIELD · held-out validation set | EPE: 0.014 -> 0.013 | Improved |
| increase latent size from 64 to 256 (ablation) | GARFIELD · held-out validation set OOD (90% conditioning) | EPE: 0.008 -> 0.019 | Hurt |
| increase latent size from 16 to 256 (ablation) | GARFIELD · held-out validation set (1% conditioning) | EPE: 0.012 -> 0.012 | No clear effect |
| GARFIELD (16 goals) vs best video model LTX | LTX · OpenVid-1M motion planning | EPE: 0.025 -> 0.013 | Improved |
| GARFIELD (4 goals) vs FPT (4 goals) | FPT · OpenVid-1M motion planning | EPE: 0.029 -> 0.020 | Improved |
| GARFIELD vs LTX sampling latency | LTX · OpenVid-1M motion planning | Latency [s]: 39 -> 0.40 | Improved |
| add entropy-based selection of conditioning points instead of an error oracle | GARFIELD · interactive conditioning | EPE | Improved |
| fine-tune GARFIELD on pedestrian data | GARFIELD 0-shot · ETH | ADE: 1.09 -> 0.84 | Improved |
| fine-tune GARFIELD on pedestrian data | GARFIELD 0-shot · HOTEL | ADE: 0.41 -> 0.29 | Improved |
Closest papers by meaning
- Envisioning the Future, One Step at a Time (CVPR 2026 · similarity 0.91)
- An Uncertain Future: Forecasting from Static Images Using Variational Autoencoders (ECCV 2016 · similarity 0.90)
- What Happens Next? Anticipating Future Motion by Generating Point Trajectories (ICLR 2026 · similarity 0.89)
- MolmoMotion: Forecasting Point Trajectories in 3D with Language Instruction (NeurIPS 2026 · similarity 0.87)
- SAMPO: Scale-wise Autoregression with Motion Prompt for Generative World Models (NeurIPS 2025 · similarity 0.87)
- PlaySlot: Learning Inverse Latent Dynamics for Controllable Object-Centric Video Prediction and Planning (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.
