Light4Mars：A Lightweight Transformer Model for Semantic Segmentation on Unstructured Environment Like Mars

Introduction

This repository is the code implementation of the paper Light4Mars：A Lightweight Transformer Model for Semantic Segmentation on Unstructured Environment Like Mars, which is based on the MMSegmentation project.

Installation

Step 0: Create a virtual environment named light4mars and activate it.

conda create -n light4mars python=3.8 -y
conda activate light4mars

Step 1: Install PyTorch 2.0.1 and torchvision 0.15.2.

pip install torch==2.0.1
pip install torchvision==0.15.2

Step 2: Install MMCV and mmsegmentation.

pip install -U openmim
mim install mmengine==0.8.4
mim install mmcv=2.0.0
pip install mmsegmentation=1.1.1

Dataset Preparation

The dataset used in the paper is SynMars-TW, which is subset of the open source unstructured environmental fine-grained synthetic dataset SynMars based on real data from the TianWen-1 mission. Please download the SynMars-TW dataset and set it according to the MMSegmentation data format.

Available datasets

Name	Size(resolution)	Object	Type	View angle	Bsed Mission
MarsData	8,390 (512*512)	Rock	semantic	Rover	Curiosity rover
MarsScapes	195(Panorama,3779 subimages )	All terrain	semantic	Rover	Curiosity rover
SynMars	60,000(1024*1024)	Rock	semantic	Rover	TianWen-1
SynMars-TW	21,000(512*512)	All terrain	depth, semantic	Rover	TianWen-1
SynMars-Air	11,700(512*512)	All terrain	semantic	UAV	TianWen-1

Model Training

python train.py configs/light4mars/light4mars-b_synmars-tw.py

Model Testing

python test.py configs/light4mars/light4mars-b_synmars-tw.py

Citation

If you use the code or performance benchmarks of this project in your research, please refer to the following bibtex citation of Light4Mars.

@article{xiong2024light4mars,
  title={Light4Mars: A lightweight transformer model for semantic segmentation on unstructured environment like Mars},
  author={Xiong, Yonggang and Xiao, Xueming and Yao, Meibao and Cui, Hutao and Fu, Yuegang},
  journal={ISPRS Journal of Photogrammetry and Remote Sensing},
  volume={214},
  pages={167--178},
  year={2024},
  publisher={Elsevier}
}