Research map / Video / Video segmentation and tracking
Envisioning the Future, One Step at a Time
Stefan Andreas Baumann, Jannik Wiese, Tommaso Martorella, Mahdi M. Kalayeh, Björn 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 |
|---|---|---|---|
| use flow-matching head instead of GMM head (ablation) | Myriad · OWM | Best-5 error: 0.110 -> 0.033 | Improved |
| add scale cascade to FM head (ablation) | Myriad · OWM | Best-5 error: 0.033 -> 0.029 | Improved |
| use FM head instead of GMM head (ablation) | Myriad · Billiard planning | accuracy: 24% -> 78% | Improved |
| model full posterior instead of regressing next step (ablation) | Myriad · Billiard planning | accuracy: 36% -> 78% | Improved |
| unroll step by step instead of full-trajectory diffusion | Myriad · Billiard planning | accuracy: 8% -> 78% | Improved |
| unroll step by step instead of single-step final-state leap (Flow Poke Transformer) | Myriad · Billiard planning | accuracy: 4% -> 78% | Improved |
| use sparse trajectories instead of dense video | Myriad · Billiard planning | accuracy: 16% -> 78% | Improved |
| increase function evaluations from 1 to 50 (ablation) | Myriad · Billiard | mean-best-of-5 EPE: 0.00361 -> 0.00138 | Improved |
| increase function evaluations from 10 to 50 (ablation) | Myriad · Billiard | mean-best-of-5 EPE: 0.00140 -> 0.00138 | No clear effect |
| random trajectory ID embedding instead of none (ablation) | Myriad · Billiard (16 trajectories) | mean-best-of-5 EPE: 0.00150 -> 0.00141 | Improved |
| random trajectory ID embedding instead of learnable (ablation) | Myriad · Billiard (16 trajectories) | mean-best-of-5 EPE: 0.00149 -> 0.00141 | Improved |
| use smaller step size dt 0.01s (50 steps) instead of 0.5s single step (ablation) | Myriad · Billiard 0.5s rollout | mean-best-of-5 EPE: 0.02823 -> 0.00141 | Improved |
Closest papers by meaning
- Schroedinger’s Cat: Probabilistic Representation and Prediction of Potential Scene Kinematics (ECCV 2026 · similarity 0.91)
- What Happens Next? Anticipating Future Motion by Generating Point Trajectories (ICLR 2026 · similarity 0.91)
- SAMPO: Scale-wise Autoregression with Motion Prompt for Generative World Models (NeurIPS 2025 · similarity 0.88)
- E-Motion: Future Motion Simulation via Event Sequence Diffusion (NeurIPS 2024 · similarity 0.88)
- Video-Mirai: Autoregressive Video Diffusion Models Need Foresight (NeurIPS 2026 · similarity 0.87)
- MESSENGER: Memory-Enhanced Sequential Scene Flow Estimation via Autoregressive Next-Frame Forecasting (NeurIPS 2026 · similarity 0.87)
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.
