A Lightweight Road Detection Algorithm Based on Multiscale Convolutional Attention Network and Coupled Decoder Head (IEEE GRSL 2023)

Aerial Remote Sensing Image Cascaded Road Detection Network Based on Edge Sensing Module and Attention Module (IEEE GRSL 2022)

This repository contains python implementation of our paper LRDNet and CasEANet.

1. Required environments:

• Linux
• Python 3.6+
• PyTorch 1.3+
• skimage

2. Use LRDNet：

Download the dataset and Pretrained Model of LRDNet

Baidu Pan：https://pan.baidu.com/s/1HT0RzsF-zMVhYuq0Bce0tg?pwd=5ous code：5ous

config file:

The config file in /configs/LRDNet_RNBD.yml is the model of LRDNet.

The config file in configs\CaeEANet_RNBD.yml is the model of CasEANet.

train the model：train.py

python train.py --config ... --work-dir ... --log_interval ... --save_every ... --resume_from ... --device ...

analysisi the complexity：analysisi_model.py

python analysisi_model.py --config ... --size ...

predict a patch image ：demo.py

python demo.py --config ... --checkpoints ... --path_img ...

predict a large image：predict.py

python predict.py --config ... --checkpoints ... --path_img ... --path_save ... --device 

browser_dataset：browser_dataset.py

python browser_dataset.py --config ... --wait_time ...

3. Visual Result

4. Citation

Please cite our paper if you find the work useful:

@ARTICLE{10098603,
  author={Liu, Dongyang and Zhang, Junping and Qi, Yunxiao and Zhang, Ye},
  journal={IEEE Geoscience and Remote Sensing Letters}, 
  title={A Lightweight Road Detection Algorithm Based on Multiscale Convolutional Attention Network and Coupled Decoder Head}, 
  year={2023},
  volume={20},
  number={},
  pages={1-5},
  doi={10.1109/LGRS.2023.3266054}}

@ARTICLE{9829847,
  author={Liu, Dongyang and Zhang, Junping and Liu, Kun and Zhang, Ye},
  journal={IEEE Geoscience and Remote Sensing Letters}, 
  title={Aerial Remote Sensing Image Cascaded Road Detection Network Based on Edge Sensing Module and Attention Module}, 
  year={2022},
  volume={19},
  number={},
  pages={1-5},
  doi={10.1109/LGRS.2022.3190495}}