Data

Tools

Indexes

Applications

Experiments

Open Source Science

legnet

[arXiv 2025] LEGNet: Lightweight Edge-Gaussian Driven Network for Low-Quality Remote Sensing Image Object Detection

https://github.com/lwcver/legnet

Science Score: 54.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • ✓
    CITATION.cff file
    Found CITATION.cff file
  • ✓
    codemeta.json file
    Found codemeta.json file
  • ✓
    .zenodo.json file
    Found .zenodo.json file
  • ○
    DOI references
  • ✓
    Academic publication links
    Links to: arxiv.org
  • ○
    Academic email domains
  • ○
    Institutional organization owner
  • ○
    JOSS paper metadata
  • ○
    Scientific vocabulary similarity
    Low similarity (6.0%) to scientific vocabulary
Last synced: 10 months ago · JSON representation ·

Repository

[arXiv 2025] LEGNet: Lightweight Edge-Gaussian Driven Network for Low-Quality Remote Sensing Image Object Detection

Basic Info
  • Host: GitHub
  • Owner: lwCVer
  • License: other
  • Language: Python
  • Default Branch: main
  • Homepage:
  • Size: 10.2 MB
Statistics
  • Stars: 55
  • Watchers: 1
  • Forks: 2
  • Open Issues: 2
  • Releases: 2
Created over 1 year ago · Last pushed about 1 year ago
Metadata Files
Readme License Citation

README.md

LEGNet: Lightweight Edge-Gaussian Driven Network for Low-Quality Remote Sensing Image Object Detection

This is the official Pytorch/Pytorch implementation of the paper:

LEGNet: Lightweight Edge-Gaussian Driven Network for Low-Quality Remote Sensing Image Object Detection

Wei Lu, Si-Bao Chen*, Hui-Dong Li, Qing-Ling Shu, Chris H. Q. Ding, Jin Tang, and Bin Luo, Senior Member, IEEE

ICCVW 2025. arXiv

Illustration of LEGNet architecture.

PWC PWC PWC PWC PWC

News 🆕

  • 2025.07.11 Congratulations! Our paper "LEGNet: A Lightweight Edge-Gaussian Network for Low-Quality Remote Sensing Image Object Detection" has been accepted by ICCVW 2025. 🔥

  • 2025.06.02 Update LEGNet V2-version paper in Arxiv. The new code, models and results are uploaded. 🎈

  • 2025.03.18 Update LEGNet original-version paper in Arxiv. The new code, models and results are uploaded. 🎈

Abstract Remote sensing object detection (RSOD) faces formidable challenges in complex visual environments. Aerial and satellite images inherently suffer from limitations such as low spatial resolution, sensor noise, blurred objects, low-light degradation, and partial occlusions. These degradation factors collectively compromise the feature discriminability in detection models, resulting in three key issues: (1) reduced contrast that hampers foreground-background separation, (2) structural discontinuities in edge representations, and (3) ambiguous feature responses caused by variations in illumination. These collectively weaken model robustness and deployment feasibility. To address these challenges, we propose LEGNet, a lightweight network that incorporates a novel edge-Gaussian aggregation (EGA) module specifically designed for low-quality remote sensing images. Our key innovation lies in the synergistic integration of Scharr operator-based edge priors with uncertainty-aware Gaussian modeling: (a) The orientation-aware Scharr filters preserve high-frequency edge details with rotational invariance; (b) The uncertainty-aware Gaussian layers probabilistically refine lowconfidence features through variance estimation. This design enables precision enhancement while maintaining architectural simplicity. Comprehensive evaluations across four RSOD benchmarks (DOTA-v1.0, v1.5, DIOR-R, FAIR1M-v1.0) and a UAVview dataset (VisDrone2019) demonstrate significant improvements. LEGNet achieves state-of-the-art performance across five benchmark datasets while ensuring computational efficiency, making it well-suited for deployment on resource-constrained edge devices in real-world remote sensing applications.

Introduction

The master branch is built on MMRotate which works with PyTorch 1.6+.

LEGNet backbone code is placed under mmrotate/models/backbones/, and the train/test configure files are placed under configs/legnet/

Pretrained Weights of Backbones

Imagenet 300-epoch pre-trained LEGNet-Tiny backbone: Download

Imagenet 300-epoch pre-trained LEGNet-Small backbone: Download

Results and Models

DOTA1.0

| Model | mAP | Angle | training mode | Batch Size | Configs | Download | |:--------------------------:|:-----:| :---: |---------------|:----------:|:--------------------------------------------------------------------------------:|:-----------------------------------------------------------------------------------------------------------------------------------:| | LEGNet-Tiny (1024,1024,200) | 79.37 | le90 | single-scale | 2*4 | orcnnlegnettinydota10testsse36.py | model | | LEGNet-Small (1024,1024,200) | 80.03 | le90 | single-scale | 2*4 | orcnnlegnetsmalldota10testsse36.py | model |

DOTA1.5

| Model | mAP | Angle | training mode | Batch Size | Configs | Download | | :----------------------: |:-----:| :---: |---| :------: |:------------------------------------------------------------------------------------------------:|:----------------------------------------------------------------------------------------------------------------:| | LEGNet-Small (1024,1024,200) | 72.89 | le90 | single-scale | 2*4 | orcnnlegnetsmalldota10testsse36.py | model |

FAIR-v1.0

| Model | mAP | Angle | training mode | Batch Size | Configs | Download | | :----------------------: |:-----:| :---: |---| :------: |:------------------------------------------------------------------------------------------------:|:----------------------------------------------------------------------------------------------------------------:| | LEGNet-Small (1024,1024,500) | 48.35 | le90 | multi-scale | 2*4 | orcnnlegnetsmallfairv1testmse12.py | model |

DIOR-R (Based on mmdetection)

| Model | mAP | Batch Size | | :------------------------------------------: |:-----:| :--------: | | LEGNet-Small | 68.40 | 1*8 |

Installation

MMRotate depends on PyTorch, MMCV and MMDetection. Below are quick steps for installation. Please refer to Install Guide for more detailed instruction.

shell conda create -n LEGNet-Det python=3.8 -y conda activate LEGNet-Det conda install pytorch==1.12.0 torchvision==0.13.0 torchaudio==0.12.0 cudatoolkit=11.3 -c pytorch pip install mmcv-full==1.7.2 -f https://download.openmmlab.com/mmcv/dist/cu113/torch1.12.0/index.html pip install mmdet pip install -v -e .

Get Started

Please see get_started.md for the basic usage of MMRotate. We provide colab tutorial, and other tutorials for:

Star History

Star History Chart

Acknowledgement

This repository is built using the timm and mmrotate repositories. MMRotate is an open source project that is contributed by researchers and engineers from various colleges and companies. We appreciate all the contributors who implement their methods or add new features, as well as users who give valuable feedbacks. We wish that the toolbox and benchmark could serve the growing research community by providing a flexible toolkit to reimplement existing methods and develop their own new methods.

If you have any questions about this work, you can contact me.

Email: luwei_ahu@qq.com; WeChat: lw2858191255.

Your star is the power that keeps us updating github.

Citation

If LEGNet is useful or relevant to your research, please kindly recognize our contributions by citing our paper: @article{lu2025legnet, title={LEGNet: Lightweight Edge-Gaussian Driven Network for Low-Quality Remote Sensing Image Object Detection}, author={Lu, Wei and Chen, Si-Bao and Li, Hui-Dong and Shu, Qing-Ling and Ding, Chris HQ and Tang, Jin and Luo, Bin}, journal={arXiv preprint arXiv:2503.14012}, year={2025} }

License

Licensed under a Creative Commons Attribution-NonCommercial 4.0 International for Non-commercial use only. Any commercial use should get formal permission first.

Owner

  • Name: 卢维
  • Login: lwCVer
  • Kind: user

student

Citation (CITATION.cff) 

cff-version: 1.2.0
message: "If you use this software, please cite it as below."
authors:
  - name: "MMRotate Contributors"
title: "OpenMMLab rotated object detection toolbox and benchmark"
date-released: 2022-02-18
url: "https://github.com/open-mmlab/mmrotate"
license: Apache-2.0

GitHub Events

Total
  • Create event: 4
  • Release event: 2
  • Issues event: 3
  • Watch event: 72
  • Issue comment event: 7
  • Public event: 1
  • Push event: 13
  • Fork event: 4
Last Year
  • Create event: 4
  • Release event: 2
  • Issues event: 3
  • Watch event: 72
  • Issue comment event: 7
  • Public event: 1
  • Push event: 13
  • Fork event: 4

Dependencies

docker/Dockerfile docker
  • pytorch/pytorch ${PYTORCH}-cuda${CUDA}-cudnn${CUDNN}-devel build
docker/serve/Dockerfile docker
  • pytorch/pytorch ${PYTORCH}-cuda${CUDA}-cudnn${CUDNN}-devel build
requirements/build.txt pypi
  • antialiased-cnns *
  • cython *
  • numpy *
requirements/docs.txt pypi
  • docutils ==0.16.0
  • markdown >=3.4.0
  • myst-parser *
  • sphinx ==4.0.2
  • sphinx-copybutton *
  • sphinx_markdown_tables >=0.0.16
  • sphinx_rtd_theme ==0.5.2
requirements/mminstall.txt pypi
  • mmcv-full >=1.5.0
requirements/optional.txt pypi
  • imagecorruptions *
  • scikit-learn *
  • scipy *
requirements/readthedocs.txt pypi
  • e2cnn *
  • mmcv *
  • mmdet >=2.25.1,<3.0.0
  • torch *
  • torchvision *
requirements/runtime.txt pypi
  • matplotlib *
  • mmcv-full *
  • mmdet >=2.25.1,<3.0.0
  • numpy *
  • pycocotools *
  • six *
  • terminaltables *
  • torch *
requirements/tests.txt pypi
  • asynctest * test
  • codecov * test
  • coverage * test
  • cython * test
  • flake8 * test
  • interrogate * test
  • isort ==4.3.21 test
  • kwarray * test
  • matplotlib * test
  • pytest * test
  • scikit-learn * test
  • ubelt * test
  • wheel * test
  • xdoctest >=0.10.0 test
  • yapf * test
requirements.txt pypi
setup.py pypi