Scholarlink

This study introduces a groundbreaking custom Convolutional Neural Network (CNN) model for the classification of cassava leaf diseases, showcasing an unprecedented level of accuracy and efficiency in agricultural image processing. At the core of this research is the innovative use of an encoder-decoder architecture, specifically tailored for high-resolution, image-to-image translation tasks. The model’s proficiency is highlighted by its remarkable performance metrics: an overall accuracy of 99%, coupled with consistently high precision, recall, and F1-scores across various disease classes. These classes include Cassava Bacterial Blight, Cassava Brown Streak Disease, Cassava Green Mottle, Cassava Mosaic Disease, and Healthy Leaves. A critical component of the model’s success is the implementation of multiple optimizers—Stochastic Gradient Descent (SGD), RMSprop, Adam, and Nadam, each with a learning rate of 0.001. The Adam optimizer emerged as the most effective, driving the model to achieve its best performance. This research not only sets a new standard in plant disease classification using deep learning but also offers a scalable and efficient tool for enhancing crop protection strategies, particularly in the realm of cassava cultivation.

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

Beyond Traditional Methods: Cassava Syndrome Scan and Deep Learning in Leaf Disease Analysis

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Irfan Sadiq Rahat,
  • Hritwik Ghosh,
  • M. V. Sangameswar,
  • Radha Mohan Pattanayak

摘要

This study introduces a groundbreaking custom Convolutional Neural Network (CNN) model for the classification of cassava leaf diseases, showcasing an unprecedented level of accuracy and efficiency in agricultural image processing. At the core of this research is the innovative use of an encoder-decoder architecture, specifically tailored for high-resolution, image-to-image translation tasks. The model’s proficiency is highlighted by its remarkable performance metrics: an overall accuracy of 99%, coupled with consistently high precision, recall, and F1-scores across various disease classes. These classes include Cassava Bacterial Blight, Cassava Brown Streak Disease, Cassava Green Mottle, Cassava Mosaic Disease, and Healthy Leaves. A critical component of the model’s success is the implementation of multiple optimizers—Stochastic Gradient Descent (SGD), RMSprop, Adam, and Nadam, each with a learning rate of 0.001. The Adam optimizer emerged as the most effective, driving the model to achieve its best performance. This research not only sets a new standard in plant disease classification using deep learning but also offers a scalable and efficient tool for enhancing crop protection strategies, particularly in the realm of cassava cultivation.