Scribd Logo
Upload

Welcome to Scribd!

  • 37 views

    Yannick Serge Obam Akou 554 Poster

    Uploaded by

    Tanankem
    This document summarizes a study that developed two deep learning models using convolutional neural networks to classify plant diseases from images. The best performing model was MobileNet V2, which achieved 93.01% accuracy on a dataset of 54,306 images from 14 plant species and 38 disease classes. An Android application was created to make the model available to farmers for early disease detection. While accurate, the models could be improved by integrating additional plant varieties and disease images from different geographic regions.

    Copyright:

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

    Yannick Serge Obam Akou 554 Poster

    Uploaded by

    Tanankem
    37 views1 page
    This document summarizes a study that developed two deep learning models using convolutional neural networks to classify plant diseases from images. The best performing model was MobileNet V2, which achieved 93.01% accuracy on a dataset of 54,306 images from 14 plant species and 38 disease classes. An Android application was created to make the model available to farmers for early disease detection. While accurate, the models could be improved by integrating additional plant varieties and disease images from different geographic regions.

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document summarizes a study that developed two deep learning models using convolutional neural networks to classify plant diseases from images. The best performing model was MobileNet V2, which achieved 93.01% accuracy on a dataset of 54,306 images from 14 plant species and 38 disease classes. An Android application was created to make the model available to farmers for early disease detection. While accurate, the models could be improved by integrating additional plant varieties and disease images from different geographic regions.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    37 views1 page

    Yannick Serge Obam Akou 554 Poster

    Uploaded by

    Tanankem
    This document summarizes a study that developed two deep learning models using convolutional neural networks to classify plant diseases from images. The best performing model was MobileNet V2, which achieved 93.01% accuracy on a dataset of 54,306 images from 14 plant species and 38 disease classes. An Android application was created to make the model available to farmers for early disease detection. While accurate, the models could be improved by integrating additional plant varieties and disease images from different geographic regions.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    You are on page 1of 1

    PLANT DISEASE CLASSIFICATION USING DEEP LEARNING MODELS

    An end to end Android application which detects plant diseases using deep learning model

    Yannick Serge OBAM AKOU


    Mentor : Dr Amidou HALIDOU

    INTRODUCTION C. GOOGLE COLAB

    Plant diseases pose a threat to food security globally as they contribute 10– CPU Google Compute Engine backend
    16% of global crop losses each year costing an estimated US$220 billion. In RAM 12.72 GB
    Africa alone, 80% of agricultural production comes from smallholders and
    DISK 358.27 GB
    the identification of plant diseases remains difficult due to the lack of neces-
    sary infrastructure. Research was carried out to realize an intelligent system D. IMPLEMENTATION PROCESS
    of detection of plant diseases based on simple images of leaves of healthy
    and diseased plants, using deep learning methods to help farmers overcome
    this problem.
    For this purpose, two models of convolutional neural networks were deve-
    loped and trained using plantVillage’s open dataset of 54,306 images of
    leaves containing 14 species, spread over 38 distinct classes of combinations
    [plants, diseases], including healthy plants.
    The best performance achieved an accuracy of 93.01%. This high success
    rate makes this model a very useful advisory or early warning tool. An An- RESULTS
    droid application was immediately developed to make it available to far-
    mers. It should be noted, however, that this approach could be further im-
    proved by integrating several other varieties of plant species and diseases
    taken under actual crop conditions and from several geographical areas.

    OBJECTIVES
    • To build and compare CNN models that classify plant diseases
    • To build an end to end AI system which detects plant diseases to
    help farmers
    • To make the project open source

    USED APPROACH
    To date, convolutional neural networks seem to
    be the most appropriate method for image-based
    plant disease classification because they are the
    most efficient.
    A. DATASET

    We used PlantVillage’s open dataset of 54 306


    images of leaves containing 14 species, spread
    over 38 distinct classes, including healthy
    plants .

    B. USED MODEL - TRANSFER LEARNING


    CONCLUSION
    Inception V3 - Preprocessing and the training Diagram Flow
    To detect plant diseases from a picture of an infected or healthy leaf, we
    developed two deep learning models based on two convolutional neural
    network architectures, namely Inception v3 and MobileNet v2.
    The best performing model architecture was MobileNet V2, which achieved
    a accuracy of 93.01% with a 0.30% loss.
    So we deployed this model in the android mobile platform.
    So a farmer in a remote location could be warned of a possible threat to his
    MobileNet V2 - Preprocessing and the training Diagram Flow
    crop and an agronomist could have a valuable advisory tool.
    Despite the high accuracy of the developed system, it is far from being a
    perfect tool that can be used under all conditions, and it will be necessary
    to integrate several other varieties of plant species and diseases from seve-
    ral sources in different geographical areas. This task will require sufficient
    financial resources and expertise. Any investment is a risk that may or may
    not be profitable: but with innovation, you always win.

    You might also like