Scholarlink

Rice has become one of the largest and most widely cultivated food crops in the world. Nevertheless, there is a significant probability that high amounts of salt, especially at the stage of seedling development, would have a detrimental effect on rice output. As a result, it is critical to quickly discover and develop salinity-tolerant rice crop types, particularly at the stage of seedling development, to prevent a decrease in rice output. Expertise from humans is required for the categorization of visual signals and the conventional way of a standard assessment system for identifying rice crop salinity stress. It's time demanding and prone to mistakes, both of which may lead to inaccurate classification. Recent developments in deep learning and transfer training have opened the road for the automated categorization of a wide range of agricultural jobs, from the detection of plant diseases and pests to the identification of weeds and invasive species. The lack of plant-domain-specific pre-trained models and the limited number of available datasets remain significant obstacles to agricultural automation. We use the AgriNet dataset, a database of 850 image graphs of farms and agricultural settings. The work shows the need for a deep learning developed model over the conventional way of measuring rice crop sensitivity to salt stress throughout the seedling stage to detect and categorize stress caused by salinity in the seedling stage of rice using field images. Thus, we use one of the most advanced deep learning approaches, VGG 16, VGG 19, Resnet 50, Resnet 100, exception, Resnet 50V2, Inception V3, MobileNetV2, DenseNet121, and ResNet101V2 which was built in Python's Jupyter Notebook, to construct the classification model. To further demonstrate the need for a computer-based classification system for salinity estimation that could be used as a tool for automating the classification process for rice development to aid scientists and farmers in the rice crop management system, the model can classify rice seedling images in accordance to the scores as Grade1, Grade3, Grade5, Grade7, and Grade9.

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

Predicting Salinity Resistance of Rice at the Seedling Stage: An Evaluation of Transfer Learning Methods

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Sharada K. Shiragudikar,
  • Geeta Bharamagoudar,
  • Manohara K. K.,
  • Malathi S. Y.,
  • Shashikumar G.Totad

摘要

Rice has become one of the largest and most widely cultivated food crops in the world. Nevertheless, there is a significant probability that high amounts of salt, especially at the stage of seedling development, would have a detrimental effect on rice output. As a result, it is critical to quickly discover and develop salinity-tolerant rice crop types, particularly at the stage of seedling development, to prevent a decrease in rice output. Expertise from humans is required for the categorization of visual signals and the conventional way of a standard assessment system for identifying rice crop salinity stress. It's time demanding and prone to mistakes, both of which may lead to inaccurate classification. Recent developments in deep learning and transfer training have opened the road for the automated categorization of a wide range of agricultural jobs, from the detection of plant diseases and pests to the identification of weeds and invasive species. The lack of plant-domain-specific pre-trained models and the limited number of available datasets remain significant obstacles to agricultural automation. We use the AgriNet dataset, a database of 850 image graphs of farms and agricultural settings. The work shows the need for a deep learning developed model over the conventional way of measuring rice crop sensitivity to salt stress throughout the seedling stage to detect and categorize stress caused by salinity in the seedling stage of rice using field images. Thus, we use one of the most advanced deep learning approaches, VGG 16, VGG 19, Resnet 50, Resnet 100, exception, Resnet 50V2, Inception V3, MobileNetV2, DenseNet121, and ResNet101V2 which was built in Python's Jupyter Notebook, to construct the classification model. To further demonstrate the need for a computer-based classification system for salinity estimation that could be used as a tool for automating the classification process for rice development to aid scientists and farmers in the rice crop management system, the model can classify rice seedling images in accordance to the scores as Grade1, Grade3, Grade5, Grade7, and Grade9.