Scholarlink

This paper presents Phytosense, an ML-based system for the early visual classification and plant disease identification. Employing transfer learning with the MobileNetV2 architecture, we created a computationally efficient, lightweight CNN model trained on the dataset obtained from PlantVillage, which contains over eighty-seven thousand images spanning 38 disease classes. The model demonstrates strong generalization with 95% overall accuracy and high performance metrics such as a ROC-AUC score of 0.999. Designed for scalability and real-time deployment, Phytosense is tailored for integration with low-cost IoT systems, supporting use cases from small-scale gardens to extensive farmland applications.

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

Phytosense: A Lightweight System for Real-Time Plant Disease Detection

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Akash Mondal,
  • Arijeet Ghosh,
  • Sudipta Dey,
  • Sreya Ghosh,
  • Apurba Nandi,
  • Kiran Ghosh

摘要

This paper presents Phytosense, an ML-based system for the early visual classification and plant disease identification. Employing transfer learning with the MobileNetV2 architecture, we created a computationally efficient, lightweight CNN model trained on the dataset obtained from PlantVillage, which contains over eighty-seven thousand images spanning 38 disease classes. The model demonstrates strong generalization with 95% overall accuracy and high performance metrics such as a ROC-AUC score of 0.999. Designed for scalability and real-time deployment, Phytosense is tailored for integration with low-cost IoT systems, supporting use cases from small-scale gardens to extensive farmland applications.