Towards Predicting Any Human Trajectory In Context
Ryo Fujii, Hideo Saito, Ryo Hachiuma · 2025
Read Towards Predicting Any Human Trajectory In Context by Ryo Fujii, Hideo Saito, Ryo Hachiuma (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-00871.txt.
Description
Ryo Fujii, Hideo Saito, and Ryo Hachiuma teach a model to predict where a pedestrian will walk in a new scene, using examples instead of retraining, in a 2025 paper. Forecasters usually need fresh data and fine-tuning for each new place, a burden for small devices. Their framework, TrajICL, adapts in context at run time.
A selection method picks past walkers in the same scene whose paths match the target's place and motion. A second pass refines the pick using the model's own draft forecast, so long-term patterns count too. Trained on a large synthetic dataset, the system beat even fine-tuned rivals across public benchmarks, in both familiar and unseen settings. The authors see in-context learning as the practical route to adaptable forecasting.
Catalog record
| CREATOR | Ryo Fujii, Hideo Saito, Ryo Hachiuma |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 41 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-00871