Transparent Peer Review By Scholar9
Plant Leaf Disease Detection using CNN
Abstract
Unidentified plant diseases cause a considerable annual crop yield loss for India each year. Manual inspection by farmers or specialists is the conventional approach of disease detection, which can be laborious and imprecise. Many small and medium-sized farms around the world are finding it to be unfeasible. A computer-aided disease recognition model is suggested as a solution to this problem. It makes use of deep convolutional networks for the classification of leaf images. In order to identify plant diseases, VGG16 and Resnet34 CNN were suggested in this work. The three processing processes are categorization, picture reduction, and feature extraction. The convolutional layer in CNN uses a plant image to extract features. The image is resized by the pooling layer.A thick layer of classification was used for diseases. Using a sample of 14 different plants, the suggested approach is able to distinguish 38 distinct plant illnesses from their surrounding foliage. It was compared how well Resnet34 and VGG16 performed. The performance metrics used were specificity, sensitivity, and accuracy. Providing farmers with tailored advice based on soil characteristics, temperature, and humidity is beneficial.
Murali Mohana Krishna Dandu Reviewer
16 Sep 2024 02:56 PM
Approved
Relevance and Originality
The Research Article is highly relevant as it tackles the significant issue of plant disease detection, which is crucial for improving crop yield and reducing losses in agriculture. The use of computer-aided disease recognition models, specifically deep convolutional networks, is an original approach that addresses the limitations of manual inspection. By leveraging VGG16 and ResNet34 CNNs, the study introduces an innovative solution for detecting a wide range of plant diseases, which can be particularly beneficial for small and medium-sized farms facing challenges with traditional methods.
Methodology
The methodology employed in the study is sound and includes the use of deep convolutional neural networks (CNNs) for image classification. The three processing steps—categorization, image reduction, and feature extraction—are well-defined and align with standard practices in image-based disease detection. The use of VGG16 and ResNet34 for feature extraction and classification is appropriate, given their proven effectiveness in similar applications. However, a more detailed explanation of the image preprocessing steps and how they impact the performance of the CNNs would strengthen the methodology section.
Validity & Reliability
The validity of the study is supported by the use of established deep learning models and performance metrics such as specificity, sensitivity, and accuracy. The comparison between ResNet34 and VGG16 provides a reliable basis for evaluating the effectiveness of the models. To enhance reliability, the study could include a discussion on the dataset's diversity and representativeness, as well as any potential biases in the sample of 14 plants used. Additionally, validation with a larger and more diverse dataset would help in assessing the model's generalizability.
Clarity and Structure
The Research Article is generally clear and well-structured. It outlines the problem, methodology, and performance metrics in a logical manner. However, to improve clarity, the paper could benefit from including visual examples of the leaf images used for classification and a detailed explanation of the CNN architecture. Diagrams or flowcharts depicting the image processing and classification stages would also help readers better understand the approach and workflow.
Result Analysis
The result analysis effectively demonstrates the ability of the proposed model to distinguish between 38 distinct plant diseases. The use of specificity, sensitivity, and accuracy as performance metrics is appropriate and provides a clear picture of the model's effectiveness. To enhance the result analysis, the paper could include a comparison of the model's performance with existing disease detection methods and discuss any potential limitations or areas for improvement. Additionally, incorporating real-world case studies or examples where the model has been applied successfully would provide practical insights into its effectiveness and utility for farmers.
IJ Publication Publisher
Done Sir
Murali Mohana Krishna Dandu Reviewer