Scholarlink

<p>Early and accurate disease detection in rice plants reduces yield losses and ensures food security. Conventional rice disease detection techniques are based on manual observation. These conventional techniques are time-consuming, labor-intensive, and prone to inaccuracies. To address these limitations, this paper proposes a&#xa0;new approach called EffiNetX. The proposed approach integrates Inverted Residual Blocks (IRBs) and squeeze-and-excitation (SE) modules to improve feature extraction while maintaining efficiency. EffiNetX extracts optimal features from rice plant images, enabling the development of a&#xa0;compact yet highly accurate model. The proposed technique was implemented in Python and tested on a&#xa0;dataset of 5932 images of four rice disease categories. The performance of the proposed technique was compared with 12&#xa0;other rice disease detection approaches. The comparative performance results showed that the proposed EffiNetX achieved 98.90% accuracy while maintaining a&#xa0;compact size of 1.42 MB. This performance indicates the superiority of the proposed approach over existing approaches in both efficiency and predictive performance.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

EffiNetX: a Lightweight Deep Learning Approach for Accurate Rice Crop Disease Detection

  • 期刊论文

  • 发表日期：

    2025年11月6日

查看全文
  • Chatter Singh,
  • Amar Singh,
  • Shakti Kumar

摘要

Early and accurate disease detection in rice plants reduces yield losses and ensures food security. Conventional rice disease detection techniques are based on manual observation. These conventional techniques are time-consuming, labor-intensive, and prone to inaccuracies. To address these limitations, this paper proposes a new approach called EffiNetX. The proposed approach integrates Inverted Residual Blocks (IRBs) and squeeze-and-excitation (SE) modules to improve feature extraction while maintaining efficiency. EffiNetX extracts optimal features from rice plant images, enabling the development of a compact yet highly accurate model. The proposed technique was implemented in Python and tested on a dataset of 5932 images of four rice disease categories. The performance of the proposed technique was compared with 12 other rice disease detection approaches. The comparative performance results showed that the proposed EffiNetX achieved 98.90% accuracy while maintaining a compact size of 1.42 MB. This performance indicates the superiority of the proposed approach over existing approaches in both efficiency and predictive performance.