iviplab / letnet

This repository is an official PyTorch implementation of our paper “LETNet: Lightweight-Real-time-Semantic-Segmentation-Network-with-Efficient-Transformer-and-CNN”.(TITS2023)

LETNet

This repository is an official PyTorch implementation of our paper "Lightweight Real-time Semantic Segmentation Network with Efficient Transformer and CNN". Accepted by IEEE TRANSACTIONS ON INTELLIGENCE TRANSPORTATION SYSTEMS, 2023. (IF: 9.551)

Paper | Code

Installation

cuda == 10.2
Python == 3.6.4
Pytorch == 1.8.0+cu101

# clone this repository
git clone https://github.com/XU-GITHUB-curry/LETNet_Lightweight-Real-time-Semantic-Segmentation-Network-with-Efficient-Transformer-and-CNN.git

Train

# cityscapes
python train.py --dataset cityscapes --train_type train --max_epochs 1000 --lr 4.5e-2 --batch_size 5

# camvid
python train.py --dataset cityscapes --train_type train --max_epochs 1000 --lr 1e-3 --batch_size 8

Test

# cityscapes
python test.py --dataset cityscapes --checkpoint ./checkpoint/cityscapes/FBSNetbs4gpu1_train/model_1000.pth

# camvid
python test.py --dataset camvid --checkpoint ./checkpoint/camvid/FBSNetbs6gpu1_trainval/model_1000.pth

Predict

only for cityscapes dataset

python predict.py --dataset cityscapes 

Results

• Please refer to our article for more details.

Methods	Dataset	Input Size	mIoU(%)
LETNet	Cityscapes	512x1024	72.8
LETNet	CamVid	360x480	70.5

Citation

If you find this project useful for your research, please cite our paper:

@article{xu2023lightweight,
  title={Lightweight Real-Time Semantic Segmentation Network With Efficient Transformer and CNN},
  author={Xu, Guoan and Li, Juncheng and Gao, Guangwei and Lu, Huimin and Yang, Jian and Yue, Dong},
  journal={IEEE Transactions on Intelligent Transportation Systems},
  year={2023},
  publisher={IEEE}
}

Thanks && Refer

@misc{Efficient-Segmentation-Networks,
  author = {Yu Wang},
  title = {Efficient-Segmentation-Networks Pytorch Implementation},
  year = {2019},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/xiaoyufenfei/Efficient-Segmentation-Networks}},
  commit = {master}
}

For more code about lightweight real-time semantic segmentation, please refer to: https://github.com/xiaoyufenfei/Efficient-Segmentation-Networks