SEANet

Official Pytorch Code base for "Using a semantic edge-aware multi-task neural network to delineate agricultural parcels from remote sensing images"

Project

Introduction

This paper aims to develop a semantic segmentation network for agricultural parcel delineation from remote sensing images, with particular attention to extracting parcels with regularized and closed boundaries. We build a semantic edge-aware multi-task neural network, called SEANet, to enhance the extraction of local and global features regarding the edge and of thematic information regarding agricultural parcels. Compared with conventional CNNs, SEANet fully uses semantic edge features at both the local and global levels, improving the geometric accuracy of agricultural parcel delineation.

Using the code:

The code is stable while using Python 3.7.0, CUDA >=11.0

• Clone this repository:

git clone https://github.com/long123524/SEANet_torch.git
cd SEANet_torch

To install all the dependencies using conda or pip:

PyTorch
TensorboardX
OpenCV
numpy
tqdm

Preprocessing

You can use the https://github.com/long123524/BsiNet-torch/blob/main/preprocess.py to obtain contour and distance maps.

Data Format

Make sure to put the files as the following structure:

inputs
└── <train>
    ├── image
    |   ├── 001.tif
    │   ├── 002.tif
    │   ├── 003.tif
    │   ├── ...
    |
    └── mask
    |   ├── 001.tif
    |   ├── 002.tif
    |   ├── 003.tif
    |   ├── ...
    └── contour
    |   ├── 001.tif
    |   ├── 002.tif
    |   ├── 003.tif
    |   ├── ...
    └── dist_contour
    |   ├── 001.tif
    |   ├── 002.tif
    |   ├── 003.tif
    └── ├── ...

For test datasets, the same structure as the above.

Pretrained weight

VGG16: link1：https://pan.baidu.com/s/1rdIWgc6DJbxMkuD3OkIoQA code：01kq link2：https://drive.google.com/file/d/1idlA55pLM4EPxWQUvuL0kRj1SHfF32ct/view?usp=drive_link

Training and testing

1. Train the model.

python train.py

2. Evaluate.

python accuracy_evaluation.py

Citation:

Li M, Long J, Stein A, et al. Using a semantic edge-aware multi-task neural network to delineate agricultural parcels from remote sensing images. ISPRS Journal of Photogrammetry and Remote Sensing, 200: 24-40.