Scholarlink

<p>The importance of crop disease detection has implications for sustainable agriculture and global food security. The sustainable production of high-quality crops can be ensured by early disease detection of the diseases. It can also assist reduce the environmental effect of crop protection techniques, encourage targeted treatments, and avoid the use of chemical pesticides. Agriculture has benefited from advances in deep neural networks as a result of technological advancement. Although deep neural networks perform well in image classification tasks, one of their drawbacks is their difficulty in decoding the results. This study looks at seven distinct infectious diseases on mango leaves. The proposed method consists of identifying and locating affected region of interest in crops using pre-trained deep learning models. Using a dataset of 4000 images (both healthy and diseased) from MangoLeafDB related to mango leaf disease, we deployed EfficientNetB5. The model’s average training accuracy was 99.84%, its validation accuracy was 98.28%, and its testing accuracy was 99.07% achieved over 10 cross folds. To improve model interpretability, gradient-weighted class activation mapping (GradCAM) and locally interpretable model-agnostic explanations (LIME) are suggested for identifying further disease probability.</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

Unveiling CNN-XAI approach in Mango health assessment

  • 期刊论文

  • 发表日期：

    2026年3月11日

查看全文
  • Vinita Chauhan,
  • Suma Dawn

摘要

The importance of crop disease detection has implications for sustainable agriculture and global food security. The sustainable production of high-quality crops can be ensured by early disease detection of the diseases. It can also assist reduce the environmental effect of crop protection techniques, encourage targeted treatments, and avoid the use of chemical pesticides. Agriculture has benefited from advances in deep neural networks as a result of technological advancement. Although deep neural networks perform well in image classification tasks, one of their drawbacks is their difficulty in decoding the results. This study looks at seven distinct infectious diseases on mango leaves. The proposed method consists of identifying and locating affected region of interest in crops using pre-trained deep learning models. Using a dataset of 4000 images (both healthy and diseased) from MangoLeafDB related to mango leaf disease, we deployed EfficientNetB5. The model’s average training accuracy was 99.84%, its validation accuracy was 98.28%, and its testing accuracy was 99.07% achieved over 10 cross folds. To improve model interpretability, gradient-weighted class activation mapping (GradCAM) and locally interpretable model-agnostic explanations (LIME) are suggested for identifying further disease probability.