Scribd Logo
Upload

Welcome to Scribd!

  • 13 views

    Plant Leaf Disease Detection Using Mask RCNN: 1. Abstract

    Uploaded by

    Adarsh Chauhan
    This paper provides an overview of using Mask R-CNN for plant leaf disease detection. Mask R-CNN is well-suited for this task as it can precisely identify and segment diseased areas of leaves. The paper surveys different methods and case studies applying Mask R-CNN to detect diseases using visible light and multispectral images. Emerging trends are discussed like leveraging pre-trained models, crowdsourcing data collection, and deploying models at the edge for real-time detection. Mask R-CNN is compared to other approaches like Faster R-CNN and traditional methods.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd

    Plant Leaf Disease Detection Using Mask RCNN: 1. Abstract

    Uploaded by

    Adarsh Chauhan
    13 views3 pages
    This paper provides an overview of using Mask R-CNN for plant leaf disease detection. Mask R-CNN is well-suited for this task as it can precisely identify and segment diseased areas of leaves. The paper surveys different methods and case studies applying Mask R-CNN to detect diseases using visible light and multispectral images. Emerging trends are discussed like leveraging pre-trained models, crowdsourcing data collection, and deploying models at the edge for real-time detection. Mask R-CNN is compared to other approaches like Faster R-CNN and traditional methods.

    Original Title

    ML

    Copyright

    © © All Rights Reserved

    Available Formats

    DOCX, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This paper provides an overview of using Mask R-CNN for plant leaf disease detection. Mask R-CNN is well-suited for this task as it can precisely identify and segment diseased areas of leaves. The paper surveys different methods and case studies applying Mask R-CNN to detect diseases using visible light and multispectral images. Emerging trends are discussed like leveraging pre-trained models, crowdsourcing data collection, and deploying models at the edge for real-time detection. Mask R-CNN is compared to other approaches like Faster R-CNN and traditional methods.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd
    Download as docx, pdf, or txt
    13 views3 pages

    Plant Leaf Disease Detection Using Mask RCNN: 1. Abstract

    Uploaded by

    Adarsh Chauhan
    This paper provides an overview of using Mask R-CNN for plant leaf disease detection. Mask R-CNN is well-suited for this task as it can precisely identify and segment diseased areas of leaves. The paper surveys different methods and case studies applying Mask R-CNN to detect diseases using visible light and multispectral images. Emerging trends are discussed like leveraging pre-trained models, crowdsourcing data collection, and deploying models at the edge for real-time detection. Mask R-CNN is compared to other approaches like Faster R-CNN and traditional methods.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd
    Download as docx, pdf, or txt
    of 3

    Plant Leaf Disease detection using Mask RCNN

    Adarsh Chauhan, Btech in Computer Science

    Graphic Era Hill University, Dehradun

    1. Abstract

    Deep learning is a contemporary and highly effective approach in image processing, particularly for
    accurate results. Numerous techniques in deep learning and image processing play a pivotal role in
    the detection and classification of leaf diseases. This includes deep learning techniques like
    Convolutional Neural Networks (CNN), Fast RCNN, Faster RCNN, and Mask RCNN, as well as image
    processing techniques encompassing image preprocessing, segmentation, and feature extraction.

    Research reveals that deep learning techniques consistently outperform traditional image processing
    methods in terms of accuracy. Plant leaf disease detection finds extensive applications in diverse
    fields, notably in biological research and agriculture. Given the substantial impact of agricultural
    productivity on the economy, the significance of this research is underscored.

    This paper offers an encompassing overview of the various methods employed in Plant Leaf Disease
    Detection. It also provides a survey of diverse disease classification techniques suitable for plant leaf
    disease detection. Several authors have explored various methods for identifying leaf diseases and
    have suggested potential implementations.

    KEYWORDS: Leaf Disease Detection, Deep Learning, Image Processing, Feature Extraction,
    Convolutional Neural Network, Mask R-CNN.

    2. Introduction

    Human health depends on the type of food we absorb and if the crops are not good, many health
    problems can arise and affect people. Therefore, good health is only possible if the crops are healthy.
    Any plant with disease will produce products that are harmful to health. Therefore, paying attention to
    leaf diseases is the most important and beneficial for a problem-free production. However, manual
    identification of leaf diseases takes a long time and is not very accurate. Therefore, advanced
    technology is now available. Advanced "Artificial Intelligence" technologies such as deep learning
    can predict leaf diseases. Disease in infected leaves cannot be predicted by human observation.
    This system is designed to support and help farmers protect their products from all kinds of pests and
    diseases without damaging agriculture and other products in the field. It aims to identify the types of
    pests and diseases on leaves and will use clever strategies to cover the number of diseases living on
    leaves. Agriculture is now turning into “digital agriculture”. Repeated practice and experiments have
    proven that with the advancement of technology, the production of quality agricultural products will
    increase.
    This process is accomplished by completing several important steps: creating data for disease
    prediction, generating text, progress (page view), performing train season metal test segmentation,
    training data and creating CNN to predict epidemic type. illness. Then create a configuration file that
    will encapsulate and process the annotation file. Transfer the image to the masking parameters of the
    Mask R-CNN algorithm to obtain the region and mask of the affected area.
    Technologies Used
    Our research uses the following main technologies:
    1. Deep Learning: Deep learning forms the basis of our approach. Using deep convolutional neural
    networks (CNN), we extract important features from leaf images that will be useful in disease
    prediction.
    2. Mask-RCNN: The basis of our detection system is Mask-RCNN. This innovation excels in sample
    segmentation and classification, making it ideal for identifying and identifying diseased areas of
    vegetation.
    3. Python Programming: The implementation of our model is done by Python, a versatile
    programming language that provides a rich ecosystem of libraries and frameworks for deep learning
    and graphics. 4. Scikit-Image: Scikit-Image is a Python library used for image processing such as
    image loading, manipulation and image removal from web pages.
    5. TensorFlow: TensorFlow is an open source machine learning and platform for us to build and train
    the Mask-RCNN model. The ability to take advantage of GPU acceleration is especially useful for
    deep learning.
    6. Data annotation tools: To organize our data, we use data annotation tools to easily plot regions of
    interest (ROI) in leaf plots. This step is necessary to train our model to correctly identify and classify
    infectious diseases.

    3. Literature Survey
    Overview of Mask R-CNN:
    Mask R-CNN is an advance Deep learning model used for object detection as well as instances
    segmentation. Mask R-CNN provides accurate object localizations, segmentations masks, and end-to-
    end training. This model has gained widespread recognition in the field of computer vision due to its
    ability to precisely identify and delineate objects within images.
    Introduction to Mask R-CNN in Agriculture:
    In recent years, Mask R-CNN has emerged as a powerful tool in the field of agriculture, enabling the
    automated detection and segmentations of plant disease. This research paper focus on the application
    of Mask R-CNN to address the critical issue of plant disease identifications. This technology has the
    potential to revolutionise disease monitoring in agriculture by providing detailed, pixel-level disease
    segmentations on plant leaves, leading to more effective disease management and crop protection. In
    this paper, I present a review of the utilization of Mask R-CNN in agriculture and discuss its
    implications and contributions to our ongoing research in the domain of plant leaf disease detection.
    Case Studies and Experiments:
    Research Project 1: Custom Dataset for Fruit Trees
    Dataset: In this project, a custom dataset of images from fruit trees in a specific region is collected.
    The dataset includes images of leaves affected by local diseases.
    Model Architecture: Mask R-CNN is configured with a VGG16 backbone to suit the project's
    computational constraints. The model is trained from scratch due to the uniqueness of the dataset.
    Performance Metrics: The project measures the model's performance by IoU, achieving an average
    score of 0.78. This model has a specific focus on a localized set of diseases.
    Research Paper 2: Multispectral Imaging in Crop Disease Detection
    Dataset: This study combines visible light images with multispectral data to enhance disease
    detection. The dataset includes both visible and multispectral images of crops.
    Model Architecture: Mask R-CNN is used with a custom-designed multispectral neural network as the
    backbone. The model leverages both types of data for improved accuracy.
    Performance Metrics: The research paper presents results that indicate a significant boost in disease
    detection accuracy compared to using visible light images alone.
    Research Project 3: Real-Time Disease Detection
    Dataset: A collection of real-time images of crops is used, captured by drones equipped with cameras.
    This dataset is continuously updated and contains dynamic environmental conditions.
    Model Architecture: Mask R-CNN is adapted to real-time processing and optimized for efficiency. A
    combination of lightweight architectures and distributed computing enables real-time disease
    detection.
    Performance Metrics: The project evaluates the model's performance in real-world, dynamic
    conditions, measuring both accuracy and processing speed. The system is capable of providing instant
    disease detection on the field.

    Emerging Trends in Plant Disease Detection:

     Transfer Learning: Leveraging pre-trained models for better results.


     Data Augmentation: Enhancing model robustness with augmented datasets.
     Spectral Data Integration: Combining spectral and RGB data for improved insights.
     Real-Time Detection: Enabling on-the-spot disease diagnosis in the field.
     Crowdsourcing and Mobile Apps: Involving farmers in data collection.
     AI on the Edge: Deploying AI models on edge devices for quick decisions.
     Interdisciplinary Research: Collaboration between experts for holistic solutions.
     Disease Forecasting: Predicting disease spread for preventive measures.

    Comparison with Other Approaches:


    1. Mask R-CNN:
    Strengths: Accurate pixel-level disease segmentation, object detection, adaptability.
    Weaknesses: Computational intensity, complexity, data dependency.
    2. Faster R-CNN:
    Strengths: Efficient object detection, speed, transfer learning.
    Weaknesses: Lacks pixel-level segmentation, requires additional segmentation for fine gained
    tasks.
    3. Traditional Methods:
    Strengths: Simplicity, real-time processing, suitable for limited data.
    Weaknesses: Limited generalization, lower accuracy, manual feature engineering.

    You might also like