Scholarlink

Crop diseases pose a serious threat to food security and livelihoods of farmers in the agriculture sector, which forms a significant portion of global economies. This research will propose a deep learning-based approach for efficient crop disease detection and management using the PlantVillage dataset of 11,400 images. Three models were considered, namely Custom CNN, VGG16, and ResNet50, and it was found that ResNet50 achieved the highest accuracy of 97.18%, outperforming VGG16 with 97.01% and Custom CNN at 90.88%. The designed system classified leaves as healthy or diseased by identifying exact infections, allowing timely interventions. ResNet50 led in terms of scalability and generalization, significantly better than traditional models. Integrated into the smart web application, the presented system allows one to detect in real time various diseases and therefore encourages precision practice in agriculture for scaling across many crops. As such, significant reduction of wastes, minimization of economic loses, and maximum productivity are found in agriculture—all factors constituting sustainable agriculture. This high power model therefore addresses some of the most important current challenges in crop health monitoring and provides straightforward applications toward great global agricultural progress.

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

Deep Convolutional Neural Networks for Accurate Crop Disease Classification

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Gopal Dutta,
  • Deepta Roy,
  • Susovan Das,
  • Kausik Naskar,
  • Neepan Biswas,
  • Md Ashifuddin Mondal,
  • Subhram Das,
  • Papri Ghosh

摘要

Crop diseases pose a serious threat to food security and livelihoods of farmers in the agriculture sector, which forms a significant portion of global economies. This research will propose a deep learning-based approach for efficient crop disease detection and management using the PlantVillage dataset of 11,400 images. Three models were considered, namely Custom CNN, VGG16, and ResNet50, and it was found that ResNet50 achieved the highest accuracy of 97.18%, outperforming VGG16 with 97.01% and Custom CNN at 90.88%. The designed system classified leaves as healthy or diseased by identifying exact infections, allowing timely interventions. ResNet50 led in terms of scalability and generalization, significantly better than traditional models. Integrated into the smart web application, the presented system allows one to detect in real time various diseases and therefore encourages precision practice in agriculture for scaling across many crops. As such, significant reduction of wastes, minimization of economic loses, and maximum productivity are found in agriculture—all factors constituting sustainable agriculture. This high power model therefore addresses some of the most important current challenges in crop health monitoring and provides straightforward applications toward great global agricultural progress.