FUSAR-Map

This repo holds code for Object-level Semantic Segmentation on The High-Resolution Gaofen-3 FUSAR-Map Dataset

Usage

1. Download pre-trained weights (2020 Gaofen Challenge & FUSAR-Map)

• Get weights of 2020 Gaofen Challenge in this link extraction code：fsjy
• Get weights of FUSAR-Map in this link extraction code：：elif

2. Prepare data

Please go to "FUSAR-Map" for details.

3. Environment

Please prepare an environment with python==3.6, tensorflow-gpu==1.15.

4. Train/Test

• Run the train script on your training dataset.

CUDA_VISIBLE_DEVICES=0 python train.py

• Run the test script (UNet, SegNet, DeepLabv3+ et. al)on your test dataset.

Reference

Citations

@ARTICLE{9369836,  author={X. {Shi} and S. {Fu} and J. {Chen} and F. {Wang} and F. {Xu}},  
    journal={IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing},   
    title={Object-level Semantic Segmentation on the High-Resolution Gaofen-3 FUSAR-Map Dataset},   
    year={2021},  
    volume={},  
    number={},  
    pages={1-1},  
    doi={10.1109/JSTARS.2021.3063797}}