Related Experiment Video
Updated: Sep 13, 2025
08:47
Author Spotlight: UAV Remote Sensing for Efficient Invasive Plant Biomass Estimation
Published on: February 9, 2024
1.6K
Real time weed identification with enhanced mobilevit model for mobile devices
Xiaoyan Liu1, Qingru Sui2, Zhihui Chen1
1Changchun Sci-Tech University, Changchun, 130600, China.
Scientific Reports
|July 27, 2025
Summary
This study introduces a lightweight deep learning model for efficient weed identification on mobile devices. The enhanced MobileViT architecture achieves high accuracy and real-time performance, optimizing agricultural technology.
Area of Science:
- Agricultural technology
- Computer vision
- Deep learning
Background:
- Deep learning models significantly improve weed identification accuracy.
- Existing models are often too large and slow for mobile embedded systems.
- Optimization research for mobile weed identification is limited.
Purpose of the Study:
- To develop a lightweight weed identification model for mobile embedded systems.
- To balance high accuracy with real-time performance for practical applications.
- To address the limitations of large, slow convolutional neural networks (CNNs).
Main Methods:
- Proposed an enhanced MobileViT architecture for weed identification.
- Incorporated the Efficient Channel Attention (ECA) module for feature extraction.
- Utilized a multi-scale retinal enhancement algorithm with color restoration for image preprocessing.
Main Results:
- Achieved an F1 score of 98.51% for weed identification.
- Demonstrated an average identification time of 89 milliseconds per image.
- The model effectively balances high accuracy with low computational complexity.
Conclusions:
- The proposed lightweight MobileViT model is suitable for real-time weed identification on mobile devices.
- The model offers a practical solution for precision agriculture by minimizing complexity and maximizing performance.
- This research contributes to the advancement of efficient AI in agriculture.
