Scholarlink

With the increasing demand for automation in agriculture, horticulture, and landscaping, the potential for machine learning to transform the field of floral recognition is significant. This study explores the development and evaluation of various neural network models, including custom networks built from scratch, networks utilizing the libraries, and networks leveraging transfer learning with pre-trained models like VGG16. Employing the TensorFlow Flowers dataset, this study methodically evaluates each model's performance in terms of accuracy, efficiency, and scalability. The findings demonstrate that the transfer learning approach, utilizing pre-trained models, significantly outperforms other models, achieving the highest accuracy rate. This underscores the effectiveness of leveraging pre-existing models trained on diverse datasets for specific tasks like flower classification. Additionally, the study highlights the importance of optimization techniques such as data preprocessing, under-sampling for class balance, and regularization methods in enhancing model performance. The results of this research contribute valuable insights into the application of machine learning in floral recognition, offering a promising avenue for future advancements in the field.

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

An Evaluation of Deep Learning Neural Networks for Automating Flower Classification

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Geerish Suddul,
  • Nadish Ramsurrun,
  • Sandhya Armoogum,
  • Riyad Dhuny

摘要

With the increasing demand for automation in agriculture, horticulture, and landscaping, the potential for machine learning to transform the field of floral recognition is significant. This study explores the development and evaluation of various neural network models, including custom networks built from scratch, networks utilizing the libraries, and networks leveraging transfer learning with pre-trained models like VGG16. Employing the TensorFlow Flowers dataset, this study methodically evaluates each model's performance in terms of accuracy, efficiency, and scalability. The findings demonstrate that the transfer learning approach, utilizing pre-trained models, significantly outperforms other models, achieving the highest accuracy rate. This underscores the effectiveness of leveraging pre-existing models trained on diverse datasets for specific tasks like flower classification. Additionally, the study highlights the importance of optimization techniques such as data preprocessing, under-sampling for class balance, and regularization methods in enhancing model performance. The results of this research contribute valuable insights into the application of machine learning in floral recognition, offering a promising avenue for future advancements in the field.