├── images └── TurboTrain_framework.png └── README.md /images/TurboTrain_framework.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ucla-mobility/TurboTrain/HEAD/images/TurboTrain_framework.png -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # TurboTrain: Towards Efficient and Balanced Multi-Task Learning for Multi-Agent Perception and Prediction 2 | 3 | [![paper](https://img.shields.io/badge/arXiv-Paper-.svg)](https://arxiv.org/pdf/2508.04682) 4 | [![supplement](https://img.shields.io/badge/Supplementary-Material-red)](https://arxiv.org/pdf/2508.04682) 5 | 6 | [ICCV 2025] This is the official implementation of "TurboTrain: Towards Efficient and Balanced Multi-Task Learning for Multi-Agent Perception and Prediction", [Zewei Zhou*](https://zewei-zhou.github.io/), [Seth Z. Zhao*](https://sethzhao506.github.io/), [Tianhui Cai](https://www.tianhui-vicky.com/), [Zhiyu Huang](https://mczhi.github.io/), [Bolei Zhou](https://boleizhou.github.io/), [Jiaqi Ma](https://mobility-lab.seas.ucla.edu/about/) 7 | 8 | ![teaser](images/TurboTrain_framework.png) 9 | 10 | TurboTrain is the first efficient and balanced multi-task learning paradigm, comprising task-agnostic self-supervised pretraining and multi-task balancing, which eliminates the need for manually designing and tuning complex multi-stage training pipelines, reducing training time, and improving performance. 11 | 12 | ## News 13 | - **`2025/08`**: [TurboTrain](https://arxiv.org/pdf/2508.04682) paper release 14 | - **`2025/06`**: [TurboTrain](https://arxiv.org/pdf/2508.04682) is accepted by [ICCV 2025](https://iccv.thecvf.com/)! 15 | 16 | ## Release Plan 17 | - **`2025/08`**: ✅ [TurboTrain](https://arxiv.org/pdf/2508.04682) paper 18 | - **`2025/12`**: Full Codebase Release. 19 | 20 | ## Acknowledgement 21 | The codebase is built upon [V2XPnP](https://github.com/Zewei-Zhou/V2XPnP) in the OpenCDA ecosystem family. 22 | 23 | 24 | ## Citation 25 | If you find this repository useful for your research, please consider giving us a star 🌟 and citing our paper. 26 | ```bibtex 27 | @inproceedings{zhou2025turbotrain, 28 | title={TurboTrain: Towards efficient and balanced multi-task learning for multi-agent perception and prediction}, 29 | author={Zhou, Zewei and Zhao, Seth Z and Cai, Tianhui and Huang, Zhiyu and Zhou, Bolei and Ma, Jiaqi}, 30 | booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision}, 31 | pages={4391--4402}, 32 | year={2025} 33 | } 34 | ``` 35 | 36 | Other useful citations: 37 | ```bibtex 38 | @article{zhao2024coopre, 39 | title={CooPre: Cooperative Pretraining for V2X Cooperative Perception}, 40 | author={Zhao, Seth Z and Xiang, Hao and Xu, Chenfeng and Xia, Xin and Zhou, Bolei and Ma, Jiaqi}, 41 | journal={arXiv preprint arXiv:2408.11241}, 42 | year={2024} 43 | } 44 | 45 | @inproceedings{zhou2025v2xpnp, 46 | title={V2xpnp: Vehicle-to-everything spatio-temporal fusion for multi-agent perception and prediction}, 47 | author={Zhou, Zewei and Xiang, Hao and Zheng, Zhaoliang and Zhao, Seth Z and Lei, Mingyue and Zhang, Yun and Cai, Tianhui and Liu, Xinyi and Liu, Johnson and Bajji, Maheswari and others}, 48 | booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision}, 49 | pages={25399--25409}, 50 | year={2025} 51 | } 52 | 53 | @article{xiang2024v2xreal, 54 | title={V2X-Real: a Largs-Scale Dataset for Vehicle-to-Everything Cooperative Perception}, 55 | author={Xiang, Hao and Zheng, Zhaoliang and Xia, Xin and Xu, Runsheng and Gao, Letian and Zhou, Zewei and Han, Xu and Ji, Xinkai and Li, Mingxi and Meng, Zonglin and others}, 56 | journal={arXiv preprint arXiv:2403.16034}, 57 | year={2024} 58 | } 59 | ``` 60 | --------------------------------------------------------------------------------