TENET: Transformer Encoding Network for Effective Temporal Flow on Motion Prediction
Computer Vision and Pattern Recognition
2022-07-04 v1 Artificial Intelligence
Abstract
This technical report presents an effective method for motion prediction in autonomous driving. We develop a Transformer-based method for input encoding and trajectory prediction. Besides, we propose the Temporal Flow Header to enhance the trajectory encoding. In the end, an efficient K-means ensemble method is used. Using our Transformer network and ensemble method, we win the first place of Argoverse 2 Motion Forecasting Challenge with the state-of-the-art brier-minFDE score of 1.90.
Cite
@article{arxiv.2207.00170,
title = {TENET: Transformer Encoding Network for Effective Temporal Flow on Motion Prediction},
author = {Yuting Wang and Hangning Zhou and Zhigang Zhang and Chen Feng and Huadong Lin and Chaofei Gao and Yizhi Tang and Zhenting Zhao and Shiyu Zhang and Jie Guo and Xuefeng Wang and Ziyao Xu and Chi Zhang},
journal= {arXiv preprint arXiv:2207.00170},
year = {2022}
}