Scholarlink

The research details a thorough assessment of tuned transfer learning systems for recognizing Indian rice varieties through the examination of 4,000 balanced images split evenly between eight types including RH-10, Sharbati, Sugandha, Sona Masoori, 1121, PR-11, 1401 and 1509. The models under assessment within this study encompass VGG-16, VGG-19, ResNet50, InceptionV3, MobileNetV2, DenseNet121, EfficientNetB0 and a lightweight deep learning among them. Performance analysis demonstrated that Lightweight Deep learning delivered 91% accuracy as its highest performance rate surpassing the other models in classification speed. The findings show that transfer learning techniques can become a practical approach to recognize rice plant varieties automatically in agricultural domains. The presented study demonstrates both model optimization methods and dataset balancing techniques while providing valuable knowledge for further research in the fields of precision agriculture and food quality assessment.

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

Evaluating Transfer Learning Models for Indian Rice Classification: The Impact of a Proposed Lightweight Deep Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Arihant Jain,
  • Ajay Kumar Suwalka,
  • Awanit Kumar,
  • Vikas Somani,
  • Sheshang Degadwala,
  • Dhairya Vyas

摘要

The research details a thorough assessment of tuned transfer learning systems for recognizing Indian rice varieties through the examination of 4,000 balanced images split evenly between eight types including RH-10, Sharbati, Sugandha, Sona Masoori, 1121, PR-11, 1401 and 1509. The models under assessment within this study encompass VGG-16, VGG-19, ResNet50, InceptionV3, MobileNetV2, DenseNet121, EfficientNetB0 and a lightweight deep learning among them. Performance analysis demonstrated that Lightweight Deep learning delivered 91% accuracy as its highest performance rate surpassing the other models in classification speed. The findings show that transfer learning techniques can become a practical approach to recognize rice plant varieties automatically in agricultural domains. The presented study demonstrates both model optimization methods and dataset balancing techniques while providing valuable knowledge for further research in the fields of precision agriculture and food quality assessment.