Rethinking RGB-Event Semantic Segmentation with a Novel Bidirectional Motion-enhanced Event Representation

Zhen Yao, Xiaowen Ying, Mooi Choo Chuah.

ArXiv | Paper | Project page

This repository contains the official Pytorch implementation of training & evaluation code and the pretrained models for BRENet.

Installation

conda env create --file environment.yml
conda activate BRENet
pip install torch==1.9.1+cu111 torchvision==0.10.1+cu111 -f https://download.pytorch.org/whl/torch_stable.html
conda install -c conda-forge cudatoolkit-dev==11.1.1
pip install mmcv-full==1.3.0 -f https://download.openmmlab.com/mmcv/dist/cu111/torch1.9.1/index.html
pip install timm==0.4.12
pip install ipython
pip install einops
pip install attrs
pip install yapf==0.40.1
pip install opencv-python==4.5.1.48
cd BRENet && pip install -e . --user

Data preparation

Please follow DATASET.md to prepare the datasets.

Trained Models

We provide trained weights for our models reported in the paper. All of the models were evaluated with 1 NVIDIA RTX A5000 GPU, and can be reproduced with the evaluation script above.

Dataset	Backbone	Resolution	mIoU	Accuracy	Download Link
DDD17	MiT-B2	200*346	78.56	96.61	[Google Drive] / [OneDrive]
DSEC	MiT-B2	440*640	74.94	95.85	[Google Drive] / [OneDrive]

Evaluation

# Single-gpu testing
python tools/test.py local_configs/BRENet/brenet.b2.640x440.dsec.80k.py /path/to/checkpoint_file
python tools/test.py local_configs/BRENet/brenet.b2.346x200.ddd17.160k.py /path/to/checkpoint_file

Training

Download backbone weights of MiT-B2 pretrained on ImageNet-1K, and put it in the folder pretrained/. Download FlowNet weights: Checkpoint trained on DSEC of eRaft in and put it in the folder pretrained/.

# Single-gpu training
python tools/train.py local_configs/BRENet/brenet.b2.640x440.dsec.80k.py
python tools/train.py local_configs/BRENet/brenet.b2.346x200.ddd17.160k.py

# Multi-gpu training
./tools/dist_train.sh local_configs/BRENet/brenet.b2.640x440.dsec.80k.py <GPU_NUM>
./tools/dist_train.sh local_configs/BRENet/brenet.b2.346x200.ddd17.160k.py <GPU_NUM>

License

This repository is under the Apache-2.0 license. For commercial use, please contact with the authors.

Acknowledgement

This codebase is built based on MMSegmentation. We thank MMSegmentation for their great contributions.

Citation

Please cite our BRENet paper and other related works if you find this useful:)

@article{yao2025rethinking,
  title={Rethinking RGB-Event Semantic Segmentation with a Novel Bidirectional Motion-enhanced Event Representation},
  author={Yao, Zhen and Ying, Xiaowen and Chuah, Mooi Choo},
  journal={arXiv preprint arXiv:2505.01548},
  year={2025}
}

EVSNet: Event-guided low-light video semantic segmentation

@inproceedings{yao2025event,
  title={Event-guided low-light video semantic segmentation},
  author={Yao, Zhen and Chuah, Mooi Choo},
  booktitle={2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
  pages={3330--3341},
  year={2025},
  organization={IEEE}
}