Let Occ Flow: Self-Supervised 3D Occupancy Flow Prediction, CoRL 2024
Yili Liu*, Linzhan Mou*, Xuan Yu, Chenrui Han, Sitong Mao, Rong Xiong, Yue Wang
$\dagger$
* Equal contribution
- [2024/09/04] - Our paper has been accepted to CoRL 2024. We will release the code in this repository.
- [2024/07/18] - We released our paper on arXiv.
- We proposed Let Occ Flow, the first self-supervised method for jointly predicting 3D occupancy and occupancy flow, by integrating 2D optical flow cues into geometry and motion optimization.
- We designed a novel attention-based temporal fusion module for efficient temporal interaction. Furthermore, we proposed a flow-oriented optimization strategy to mitigate the training instability and sample imbalance problem.
- We conducted extensive experiments on various datasets with qualitative and quantitative analyses to show the competitive performance of our approach.
If this work is helpful for your research, please consider citing the following paper:
@article{liu2024letoccflow,
title={Let Occ Flow: Self-Supervised 3D Occupancy Flow Prediction},
author={Yili Liu and Linzhan Mou and Xuan Yu and Chenrui Han and Sitong Mao and Rong Xiong and Yue Wang},
journal={arXiv preprint arXiv:2407.07587},
year={2024},
}