AEGFNet

This repository provides the code for the methods and experiments presented in our paper 'Adaptive Edge-Guided Fusion Network for Remote Sensing Image Change Detection'.

Datasets

Download the building change detection dataset.

• [SYSU]
• WHU-CD
• LEVIR-CD

Prepare datasets into following structure,

├─train
    ├─A
    ├─B
    ├─label
├─val
    ├─A
    ├─B
    ├─label
├─test
    ├─A
    ├─B
    ├─label

In the following experiments, each image in the dataset is pre-cropped into multiple image patches of size 256 × 256.

Preparation

• Install dependencies using

pip install -r requirements.txt

Train

• Change the root address in the train section to your dataset address, for example: root = 'D:\dataset\LEVIR-CD'.

• Run main_train.py to start training.

• Run eval_cd.py to start testing.