Related Experiment Video
Updated: Jul 19, 2026
13:19
Deep Neural Networks for Image-Based Dietary Assessment
Published on: March 13, 2021
9.0K
Resource-optimized cnns for real-time rice disease detection with ARM cortex-M microprocessors.
Hermawan Nugroho1, Jing Xan Chew2, Sivaraman Eswaran3
1Electrical and Electronic Engineering Department, University of Nottingham Malaysia, Jln Broga, 43500, Semenyih, Malaysia. hermawan.nugroho@nottingham.edu.my.
Plant Methods
|October 16, 2024
Summary
This study uses Artificial Intelligence (AI) and Convolutional Neural Networks (CNNs) on microcontrollers for rice plant disease detection. MobileNetV2 achieved 97.5% accuracy, offering a viable smart agriculture solution.
Area of Science:
- Agricultural Science
- Computer Science
- Artificial Intelligence
Background:
- Rice is a crucial staple food, facing declining self-sufficiency in regions like Malaysia.
- Effective rice plant disease detection is vital for enhancing agricultural productivity and sustainability.
- Resource-constrained environments necessitate efficient AI models for on-site agricultural monitoring.
Purpose of the Study:
- To explore the application of Convolutional Neural Networks (CNNs) on ARM Cortex-M microprocessors for detecting rice plant diseases.
- To evaluate the performance and computational efficiency of MobileNetV2 and FD-MobileNet models for this task.
- To investigate resource optimization strategies for AI models in smart agriculture applications.
Main Methods:
- Utilized two large datasets (5,932 and 10,407 images) covering multiple rice disease classes for training and validation.
- Implemented and optimized MobileNetV2 and FD-MobileNet models for the ARM Cortex-M4 microprocessor.
- Evaluated model performance based on accuracy, computational efficiency, and resource consumption (RAM, flash memory).
Main Results:
- MobileNetV2 achieved a high accuracy of 97.5%, outperforming FD-MobileNet (90% accuracy).
- MobileNetV2 demonstrated superior performance in detecting complex diseases like tungro (93% accuracy).
- Resource optimization, even minor, significantly improved validation accuracy, highlighting the trade-off between efficiency and performance.
Conclusions:
- Deploying CNNs on microcontrollers offers a practical solution for real-time, on-site rice plant disease detection.
- The study demonstrates the potential for improved accuracy and operational efficiency in smart agriculture.
- This research contributes to addressing food security challenges through the integration of AI in agriculture.
