Scholarlink

This study investigated how neural networks could identify soil images and recommend suitable crops based on the soil. We evaluated five CNN models to assess the model’s output and performance: Sequential CNN, Sequential CNN with Data Augmentation, VGG16, Modified VGG16, and Modified VGG16 with Data Augmentation. The Sequential CNN and Modified VGG16 perform excellently in terms of accuracy and loss. Among these, the Modified VGG16 outperformed all the other models we tested. We achieved 96% training accuracy and 68% validation accuracy. In terms of testing performance, Sequential CNN and Modified VGG16 excel. Nonetheless, Sequential CNN surpassed all of the other models we tested, achieving 94% accuracy in image testing.

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 Learning-Based Soil Image Classification for Sustainable Agriculture: A Case Study on Crop Recommendation

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Md. Monwar Hossain,
  • Miskath Jahan Simu,
  • Najmus Sakib,
  • K. M. Safin Kamal,
  • Ahmed Wasif Reza

摘要

This study investigated how neural networks could identify soil images and recommend suitable crops based on the soil. We evaluated five CNN models to assess the model’s output and performance: Sequential CNN, Sequential CNN with Data Augmentation, VGG16, Modified VGG16, and Modified VGG16 with Data Augmentation. The Sequential CNN and Modified VGG16 perform excellently in terms of accuracy and loss. Among these, the Modified VGG16 outperformed all the other models we tested. We achieved 96% training accuracy and 68% validation accuracy. In terms of testing performance, Sequential CNN and Modified VGG16 excel. Nonetheless, Sequential CNN surpassed all of the other models we tested, achieving 94% accuracy in image testing.