Road-Former

Offical RoadFormer series scene parsing networks implementation based on mmsegmentation v1.0.0. All the related code and datasets will be released upon publication.

News

• [2023/09/16]: Our paper RoadFormer is submitted to the ICRA 2024.
• [2024/01/31]: Unfortunately, RoadFormer was rejected by ICRA 2024 after peer review 😭😭😭.
• [2024/03/31]: Our paper RoadFormer has been accepted by the IEEE T-IV 2024 😊😊😊. Stay tuned for the RoadFormer+!
• [2024/07/17]: Our series of works, RoadFormer and RoadFormer+, are nearing completion. Upon finalizing the organization of the code, we will release it. Stay tuned for updates!
• [2024/08/21]: Our paper RoadFormer+ has been accepted by the IEEE T-IV 2024 😊😊😊. Stay tuned for our code!
• [2024/08/22]: The implementation of our RoadFormer series, has been released, the utilized SYN-UDTIRI will be released soon.
• [2024/08/27]: Our SYN-UDTIRI dataset is now available on BaiDu Cloud, it includes the train, val, test sets of the datasets. It's about 46 GB in total, considering the test set is too large (refer to our RoadFormer paper; the test set contains 3000+ images), we plan to construct a smaller test and deploy it on our online benchmark: UDTIRI. Please refer to this site for more details about our UDTIRI series benchmark.

Citation

If you find our works useful in your research, please consider citing:

@article{li2024roadformer,
  title={RoadFormer: Duplex transformer for RGB-normal semantic road scene parsing},
  author={Li, Jiahang and Zhang, Yikang and Yun, Peng and Zhou, Guangliang and Chen, Qijun and Fan, Rui},
  journal={IEEE Transactions on Intelligent Vehicles},
  year={2024},
  publisher={IEEE},
  note={{DOI}:{10.1109/TIV.2024.3388726}},
}

@article{huang2024roadformer+,
  title={RoadFormer+: Delivering RGB-X Scene Parsing through Scale-Aware Information Decoupling and Advanced Heterogeneous Feature Fusion},
  author={Huang, Jianxin and Li, Jiahang and Jia, Ning and Sun, Yuxiang and Liu, Chengju and Chen, Qijun and Fan, Rui},
  journal={IEEE Transactions on Intelligent Vehicles},
  year={2024},
  publisher={IEEE},
  note={{DOI}:{10.1109/TIV.2024.3448251}},
}

Usage

Installation

git clone https://github.com/LiJiahang617/Road-Former.git
cd Road-Former

Please refer to MMSegmentation for the dependency instructions. You will need a compiled .so file, or you can download the pre-compiled file from here:

• _ext.cpython-38-x86_64-linux-gnu.so

This file is complied using the NVIDIA RTX 3090 GPU, which may cause error when under other devices.

Running

python tools/train.py --config <config-file-path>
python tools/test.py --config <config-file-path> --pretrained <pre-trained-pth-path>

For more training and inference details, please refer to MMSegmentation instructions.

If you find our work helpful, please consider a star⭐ for us!