Scholarlink

<p>This study presents a systematic review of artificial intelligence applications, particularly machine learning and deep learning models, for okra leaf disease detection and diagnosis. Due to the limited number of okra-specific studies, a related study on leaf diseases of other crops was included for comparative analysis. Following the PRISMA framework, 28 peer-reviewed articles published between 2015 and 2025 were examined. The results show that convolutional neural networks dominate the current study, accounting for nearly 70% of all reviewed models. Architectures such as MobileNet, ResNet, and InceptionV3 consistently achieved accuracies above 90%. Okra-focused studies reported performance ranging from 87% to 98.63% (mean = 94.5%) but were constrained by small or imbalanced datasets. In contrast, studies on other crops achieved accuracies between 77% and 99.84% (mean = 96%), supported by substantially larger datasets. The review further identifies limited adoption of explainable AI, vision transformers, and federated learning approaches. Key study gaps include dataset scale, environmental integration, real-world validation, and reproducibility. The study recommends developing large-scale okra-specific datasets, integrating agronomic variables, deploying lightweight, mobile-ready architectures, and strengthening the adoption of explainable and federated learning frameworks to enable scalable, field-deployable diagnostic systems.</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

Application of artificial intelligence for okra leaf and other plant disease detection and diagnoses: a systematic literature review

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月1日

查看全文
  • Justice Williams Asare,
  • Samuel Kotey Adjei,
  • Samuel Ampaabeng,
  • Emmanuel Akwah Kyei,
  • Ebenezer Komla Gavua,
  • Mustapha Adamu Mohammed,
  • Seth Alornyo,
  • Martin Mabeifam Ujakpa

摘要

This study presents a systematic review of artificial intelligence applications, particularly machine learning and deep learning models, for okra leaf disease detection and diagnosis. Due to the limited number of okra-specific studies, a related study on leaf diseases of other crops was included for comparative analysis. Following the PRISMA framework, 28 peer-reviewed articles published between 2015 and 2025 were examined. The results show that convolutional neural networks dominate the current study, accounting for nearly 70% of all reviewed models. Architectures such as MobileNet, ResNet, and InceptionV3 consistently achieved accuracies above 90%. Okra-focused studies reported performance ranging from 87% to 98.63% (mean = 94.5%) but were constrained by small or imbalanced datasets. In contrast, studies on other crops achieved accuracies between 77% and 99.84% (mean = 96%), supported by substantially larger datasets. The review further identifies limited adoption of explainable AI, vision transformers, and federated learning approaches. Key study gaps include dataset scale, environmental integration, real-world validation, and reproducibility. The study recommends developing large-scale okra-specific datasets, integrating agronomic variables, deploying lightweight, mobile-ready architectures, and strengthening the adoption of explainable and federated learning frameworks to enable scalable, field-deployable diagnostic systems.