This repo is the implementation of the following paper:
STrajNet: Multi-Model Hierarchical Transformer for Occupancy Flow Field Prediction in Autonomous Driving
Haochen Liu, Zhiyu Huang, Chen Lv
AutoMan Research Lab, Nanyang Technological University
[arXiv]
- This implementation is improved upon our solutions in ablations, which has won the 2nd place for 2022 Waymo Occupancy and Flow Prediction Challenge
- 🥈2022 Waymo Occupancy and Flow Prediction
- 🚗CVPR2022 Workshop on Autonomous Driving website
- 📑Waymo Occupancy and Flow Prediction Website
- Our implementation has achieved 3 SOTA metrics (Sep-2022), outperforms other baselines without using pretrained models, and requires less memory costs:
- Model Framework Overview:
Download the Waymo Open Motion Dataset; the files in tf_exmaple/ are used. Place the downloaded files into training and testing folders separately.
Also download the required scenario IDs files (.txt) for evalutation and testing
If you find this work useful please cite us 😀
@misc{https://doi.org/10.48550/arxiv.2208.00394,
doi = {10.48550/ARXIV.2208.00394},
url = {https://arxiv.org/abs/2208.00394},
author = {Liu, Haochen and Huang, Zhiyu and Lv, Chen},
keywords = {Computer Vision and Pattern Recognition (cs.CV), Artificial Intelligence (cs.AI), FOS: Computer and information sciences, FOS: Computer and information sciences},
title = {STrajNet: Multi-modal Hierarchical Transformer for Occupancy Flow Field Prediction in Autonomous Driving},
publisher = {arXiv},
year = {2022},
copyright = {Creative Commons Attribution 4.0 International}
}