Related Experiment Video
Updated: Aug 5, 2026
08:25
Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment
Published on: May 7, 2019
LFODet: Lightweight Few-Shot Object Detection with Meta-Learning in Remote Sensing Images
Haoran Wu1,2, Xuan Fang1,2, Haonan Xiong1,2
1College of Electrical Engineering, Sichuan University, Chengdu 610065, China.
Sensors (Basel, Switzerland)
|July 28, 2026
Summary
This study introduces LFODet, a lightweight few-shot object detection network for remote sensing. It balances accuracy and efficiency, enabling rapid adaptation to new targets with limited data.
Area of Science:
- Remote Sensing
- Computer Vision
- Machine Learning
Background:
- Balancing detection accuracy and model lightweightness is crucial for remote sensing object detection.
- Convolutional neural networks often require large datasets, hindering few-shot detection of novel classes.
Purpose of the Study:
- Propose LFODet, a lightweight few-shot object detection network using meta-learning.
- Enable rapid adaptation to novel classes with limited samples while maintaining base class performance.
Main Methods:
- Utilize two parallel branches for novel class adaptation.
- Integrate Semantic Ghost Channel Attention (GCA) and Fine-Grained Ghost Spatial Attention (GSA) for feature representation.
- Employ Ghost convolutions to reduce computational complexity and train in three stages: pre-training, meta-learner optimization, and fine-tuning.
Main Results:
- LFODet achieves stable and balanced performance across various few-shot learning scenarios.
- Demonstrated effectiveness on DIOR and NWPU VHR-10 benchmark datasets.
Conclusions:
- LFODet offers a practical solution for resource-constrained remote sensing applications.
- Provides rapid adaptation capabilities for new targets in remote sensing imagery.