{"ID":2888434,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2507.23657","arxiv_id":"2507.23657","title":"OmniTraj: Pre-Training on Heterogeneous Data for Adaptive and Zero-Shot Human Trajectory Prediction","abstract":"While large-scale pre-training has advanced human trajectory prediction, a critical challenge remains: zero-shot transfer to unseen dataset with varying temporal dynamics. State-of-the-art pre-trained models often require fine-tuning to adapt to new datasets with different frame rates or observation horizons, limiting their scalability and practical utility. In this work, we systematically investigate this limitation and propose a robust solution. We first demonstrate that existing data-aware discrete models struggle when transferred to new scenarios with shifted temporal setups. We then isolate the temporal generalization from dataset shift, revealing that a simple, explicit conditioning mechanism for temporal metadata is a highly effective solution. Based on this insight, we present OmniTraj, a Transformer-based model pre-trained on a large-scale, heterogeneous dataset. Our experiments show that explicitly conditioning on the frame rate enables OmniTraj to achieve state-of-the-art zero-shot transfer performance, reducing prediction error by over 70\\% in challenging cross-setup scenarios. After fine-tuning, OmniTraj achieves state-of-the-art results on four datasets, including NBA, JTA, WorldPose, and ETH-UCY. The code is publicly available: https://github.com/vita-epfl/omnitraj","short_abstract":"While large-scale pre-training has advanced human trajectory prediction, a critical challenge remains: zero-shot transfer to unseen dataset with varying temporal dynamics. State-of-the-art pre-trained models often require fine-tuning to adapt to new datasets with different frame rates or observation horizons, limiting...","url_abs":"https://arxiv.org/abs/2507.23657","url_pdf":"https://arxiv.org/pdf/2507.23657v1","authors":"[\"Yang Gao\",\"Po-Chien Luan\",\"Kaouther Messaoud\",\"Lan Feng\",\"Alexandre Alahi\"]","published":"2025-07-31T15:37:09Z","proceeding":"cs.CV","tasks":"[\"cs.CV\"]","methods":"[\"Transformer\"]","has_code":false,"code_links":[{"ID":611540,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_id":2888434,"paper_url":"https://arxiv.org/abs/2507.23657","paper_title":"OmniTraj: Pre-Training on Heterogeneous Data for Adaptive and Zero-Shot Human Trajectory Prediction","repo_url":"https://github.com/vita-epfl/omnitraj","is_official":false,"mentioned_in_paper":false,"mentioned_in_github":true,"github_stars":0}]}