PPT: Pretraining with Pseudo-Labeled Trajectories for Motion Forecasting
Yihong Xu, Yuan Yin, Éloi Zablocki, Tuan-Hung Vu, Alexandre Boulch, Matthieu Cord · 2024
Read PPT: Pretraining with Pseudo-Labeled Trajectories for Motion Forecasting by Yihong Xu, Yuan Yin, Éloi Zablocki, Tuan-Hung Vu, Alexandre Boulch, Matthieu Cord (2024) 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-2412-06491.txt.
Description
Yihong Xu and co-authors pretrain driving forecasters on labels no human ever drew. Their 2024 paper tackles motion forecasting, which guesses how road users will move. Top models learn from hand-built trajectory sets that cost dearly, scale poorly, and drift from real streets. Their frame, PPT, instead harvests raw paths from off-the-shelf detectors and trackers.
Those machine-made paths are messy, but they still teach sturdy scene sense. With a light fine-tune on a little labeled data, PPT models score well on standard tests. Gains show most in low-data, cross-domain, and end-to-end settings. The authors stress ease: the recipe is simple to add and lifts generalization for safer self-driving stacks.
Catalog record
| CREATOR | Yihong Xu, Yuan Yin, Éloi Zablocki, Tuan-Hung Vu, Alexandre Boulch, Matthieu Cord |
|---|---|
| YEAR | 2024 |
| 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/2412-06491