Adaptive Edge-aware Semantic Interaction Network (AESINet)

Official PyTorch implementation of the IEEE TGRS 2023 paper "Adaptive Edge-aware Semantic Interaction Network for Salient Object Detection in Optical Remote Sensing Images".

Authors

Xiangyu Zeng¹, Mingzhu Xu¹*, Yijun Hu¹, Haoyu Tang¹, Yupeng Hu¹, Liqiang Nie²

¹ Shandong University
² Harbin Institute of Technology (Shen Zhen)
* Corresponding author

Links

• Paper: IEEE Xplore
• Code Repository: GitHub

---

Table of Contents

• Introduction
• Highlights
• Method / Framework
• Project Structure
• Installation
• Checkpoints / Models
• Usage
• TODO
• Citation
• Contact
• License

---

Introduction

This project is the official implementation of the paper "Adaptive Edge-aware Semantic Interaction Network for Salient Object Detection in Optical Remote Sensing Images".

This project mainly addresses the challenge of Salient Object Detection (SOD) in optical remote sensing images.

• Core Idea：Enhance boundary localization and effectively fuse multi-scale semantic information by introducing an adaptive edge-aware semantic interaction mechanism.
• This Repository Provides：
  • Training and inference code based on VGG and ResNet backbones.
  • Complete pretrained models and trained checkpoint links.
  • Data processing scripts for generating dataset path list files.

Example Description

We present AESINet, a framework for Salient Object Detection (SOD) in Optical Remote Sensing Images.
Our method addresses accurate boundary localization by introducing adaptive edge-aware semantic interaction.
This repository provides the official implementation, pretrained checkpoints, and data processing scripts.

---

Highlights

• Supports salient object detection in remote sensing images.
• Provides implementations with both VGG and ResNet backbone networks.
• Includes an adaptive edge-aware module to effectively handle targets in complex remote sensing backgrounds.

---

Method / Framework

AESINet aims to improve detection performance through semantic interaction and edge-aware modules.

Framework Figure

Figure 1. Overall framework of AESINet.

---

Project Structure

.
├── assets/                # Images and framework diagrams
├── generateTrainList.py   # Script to generate training set path list
├── generateTestList.py    # Script to generate test set path list
├── main.py                # Main execution script (example)
├── models/                # AESINet model implementations (ResNet & VGG versions)
├── README.md
└── requirements.txt

---

Installation

1. Clone the repository

git clone [https://github.com/iLearn-Lab/AESINet-TGRS.git](https://github.com/iLearn-Lab/AESINet-TGRS.git)
cd AESINet-TGRS

2. Create environment

python -m venv .venv
source .venv/bin/activate   # Linux / Mac

3. Install dependencies

pip install -r requirements.txt

---

Checkpoints / Models

Model	Link	Password
AESINet-V (VGG)	Baidu Drive	qegm
AESINet-R (ResNet)	Baidu Drive	tj25
Pretrained (VGG & ResNet)	Baidu Drive	lb8l

Please place the downloaded models into the project directory and modify the corresponding path in the code.

---

Usage

1. Data Preparation

First, download the dataset and use the following scripts to generate .txt path list files:

python generateTrainList.py
python generateTestList.py

Note：Please ensure that the dataset paths are correctly configured in the code.

2. Training

python main.py --mode=train

3. Testing

python main.py --mode=test

---

TODO

• Initial release of ResNet and VGG codes.
• Release model checkpoints.
• Improve code readability and documentation.

---

Citation

@ARTICLE{AESINet2023TGRS,
  author={Zeng, Xiangyu and Xu, Mingzhu and Hu, Yijun and Tang, Haoyu and Hu, Yupeng and Nie, Liqiang},
  journal={IEEE Transactions on Geoscience and Remote Sensing}, 
  title={Adaptive Edge-Aware Semantic Interaction Network for Salient Object Detection in Optical Remote Sensing Images}, 
  year={2023},
  volume={61},
  number={},
  pages={1-16},
  doi={10.1109/TGRS.2023.3300317}}

---

Contact

If you have any questions, please contact the authors via email: z15264367990@163.com

---

License

This project is released under the Apache License 2.0.