Related Experiment Video
Updated: Jan 15, 2026
08:47
Author Spotlight: UAV Remote Sensing for Efficient Invasive Plant Biomass Estimation
Published on: February 9, 2024
2.1K
A lightweight convolutional neural network for tea leaf disease and pest recognition
Xiaojie Wen1,2, Qi Liu1,2, Xuanyuan Tang1,2
1Key Laboratory of Pest Monitoring and Safety Control of Crops and Forests of the Xinjiang Uygur Autonomous Region, College of Agronomy, Xinjiang Agricultural University, Urumqi, 830052, China.
Plant Methods
|October 15, 2025
Summary
This study introduces a new dataset and an enhanced MnasNet model for detecting tea diseases and pests. The model achieves high accuracy, enabling automated detection for precision agriculture.
Area of Science:
- Agricultural Science
- Computer Vision
- Plant Pathology
Background:
- Tea production is crucial for China's green economy, but diseases and pests threaten agricultural productivity.
- Current diagnostic methods for tea plant health suffer from limited data and poor feature representation.
- Accurate and timely identification of tea plant diseases and pests is essential for sustainable agriculture.
Purpose of the Study:
- To develop and validate an automated system for tea disease and pest detection.
- To address data scarcity and improve feature discriminability in classification models.
- To enhance precision agriculture through intelligent tea plantation management.
Main Methods:
- A new tea disease and pest dataset (TDPD) with a 23-class taxonomy was created.
- Five lightweight convolutional neural networks (LCNNs) were evaluated using various optimizers and learning rate strategies.
- An enhanced MnasNet model incorporating SimAM attention mechanisms was developed and tested.
Main Results:
- The enhanced MnasNet model achieved 98.03% accuracy on the proprietary TDPD dataset.
- The model demonstrated 84.58% accuracy on an open-access dataset.
- The SimAM attention mechanism improved feature discriminability and classification accuracy.
Conclusions:
- The developed model offers a robust solution for automated tea disease and pest detection.
- This research provides a practical framework for integrating intelligent systems into tea plantation management.
- The findings support the application of UAV-mounted imaging and mobile platforms for real-time diagnostics.
