Scholarlink

<p>Palm tree cultivation in the Al-Kharj region is significantly affected by a range of diseases that compromise both yield and crop quality. We propose a robust framework for palm tree disease classification that leverages transfer learning in combination with a Chaotic Red Panda Optimization (CRPO) algorithm. The framework effectively tackles challenges posed by small and imbalanced datasets through automated dataset exploration, class discovery, and visualization of sample distributions. The CRPO algorithm utilizes a chaotic logistic map to generate candidate fully connected layer weights, thereby optimizing the network before fine-tuning with the Adam optimization algorithm. Comprehensive visualizations illustrate the dynamics of the optimization process, including fitness evolution, population diversity, and chaotic sequences, as well as training progression and evaluation metrics. Experimental evaluation on a dataset of 3,089 images spanning nine disease categories and healthy palms demonstrates robust classification performance, with precision, recall, and F1-scores ranging from 0.98 to 1.00 across all classes.</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

Enhanced chaotic sequence population optimization for accelerated deep neural network convergence: a review and case study on palm tree classification in Al-Kharj

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月2日

查看全文
  • Zahraa Tarek,
  • Esraa Hasan

摘要

Palm tree cultivation in the Al-Kharj region is significantly affected by a range of diseases that compromise both yield and crop quality. We propose a robust framework for palm tree disease classification that leverages transfer learning in combination with a Chaotic Red Panda Optimization (CRPO) algorithm. The framework effectively tackles challenges posed by small and imbalanced datasets through automated dataset exploration, class discovery, and visualization of sample distributions. The CRPO algorithm utilizes a chaotic logistic map to generate candidate fully connected layer weights, thereby optimizing the network before fine-tuning with the Adam optimization algorithm. Comprehensive visualizations illustrate the dynamics of the optimization process, including fitness evolution, population diversity, and chaotic sequences, as well as training progression and evaluation metrics. Experimental evaluation on a dataset of 3,089 images spanning nine disease categories and healthy palms demonstrates robust classification performance, with precision, recall, and F1-scores ranging from 0.98 to 1.00 across all classes.