Final Review 2

ADIKAVI NANNAYA UNIVERSITY RAJAMAHENDRAVARAM UNIVERSITY COLLEGE OF ENGINEERING DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

Plant Disease Recognition based on Leaf Image Classification Under the Guidance of Dr. M Kamala Kumari

Presented By T V S Ravi Teja 178297601023

Index 1.

Introduction

2.

Abstract

3.

Existing and Proposed System

4.

Requirement Analysis a) Hardware requirements b) Software Requirements c) Functional and Non Functional Requirements

5.

Architecture

6.

Design a) Functional Design b) Design with UML Diagrams

7.

Implementation a) Platform and Tools b) Sample code

8.

Output

9.

Conclusion

Introduction  One of the major problems in the field of Agriculture is identification of Diseases on hand before its getting spread to a vast area.  Usually farmers or experts observe the plants with naked eye for detection and identification of disease. But this method can be time consuming , expensive and inaccurate.  By the time the disease is identified, it may to late and cause enough damage to the crop.

Abstract  To solve the above problem a tool can be developed that uses computer vison, that can classify images based on the diseases.  In order to bring it closer to the common people, this tool is developed as a mobile application, such that it can be directly used from mobile only.  This “tool” works on image processing and classifications using CNN.  For the mobile application, flutter framework is used which is capable of running on both Android as well as IOS

Existing System  In Traditional System scientists from agricultural department will come and inspect the situation, take some samples, and after testing they will suggest the appropriate cure for the problem.  If the disease is already known they is suggest the sure immediately, but if not it takes some time.  But above step is also done after the appropriate observation only, which also takes some time.  Few applications are already available but using similar procedure as mine but not using CNN.

Proposed System  A Mobile application which captures image as input processes; a model which is developed by CNN.  This model was developed by training with the images of the diseases of the crop.  After scanning the leaf, through app, its details should be displayed along with the medication.

 Whenever there is outbreak of new disease the admin should the facility to add up the details of that new disease into the application.

Requirement Analysis  Hardware Requirements: RAM:- 8GB or above (higher recommended)  Hard disk:- 1TB

 Processor:- i5 or above  Sofware Requirements: Operating System:- Windows  Work station:- Google Colab/ Anaconda

 Editor:- Vs Code  Front end:- Android Studio, Flutter, Java  Language:- Dart, Python  DataSet :- Tobacco Disease Dataset

continue…  Functional Requirements: This model is used to predict whether the Leaf which is given as input has a disease or not, if its having a disease, it will classify which type of disease it is.  The dataset which is taken as a Input, consists of all the images, separated with respective disease. On the whole the set of diseases are separated as train, test, and validation datasets to get a better output.

 Non Functional Requirements: Accuracy: finding the Accuracy for different types of classifier classes taken, such that we may know which classifier model will give a better output.

Architecture

Functional Design ➢ CNN Modules :-

a.

Redefining the selected classifier algorithm, and adding output label's as final layers

b.

Data Augmentation, applying various methods to change the positional and aspect ratio characteristics of the image, and adding to the model eg: horizontal flip, Zoom etc.

c.

Training and testing the model

d.

converting the generated model into tflite format Defining the Classifier Algorithm Data Augmentation Training the model

Tflite file

continue.. ➢ Mobile Application :a.

Load image :- either “Take Image” from camera else “Select Image” from the gallery .

b.

Classify image:- Using the model.tflite file , the image taken from above gets classified, with its respective label.

c.

Display image:- After getting a appropriate result, the image along with its respective class is displayed

UML Diagrams  UseCase Diagram:

Continue…  Class diagram :-

Continue…  Sequence Diagram:-

Implementation Neural Networks:  Neural networks(NN) are set layers of highly interconnected processing elements (neurons) that make a series of transformations on the data to generate its own understanding of it(what we commonly call features).  Neural Networks mimic the human brain, the data is stored in storage elements called neurons, and these nodes are interconnected through nodes. NN consists of different hidden layers which filtering on the previous layers, the crucial features are stored along the hidden layers.

Continue…

Neural Network with 2 hidden layers

Continue …  Convolutional Neural Network: A Convolutional Neural Network is a Deep Learning algorithm which can take in an input image, assign importance (learnable weights and biases) to various aspects/objects in the image and be able to differentiate one from the other.

Continue…  Tensor flow:  TensorFlow is a software library or framework, designed by the Google team to implement machine learning and deep learning concepts. It combines the computational algebra of optimization techniques for easy calculation of many mathematical expressions.

 Tensor Refers to the representation of t data as multi-dimensional array which represents the object of consideration whereas the term flow refers to the series of operations that one performs on tensor.

Continue…  Transfer learning : IMAGENET  It is very hard and time consuming to collect images belonging to a domain of interest and train a classifier from scratch. So, we use a pre-trained model as our base and change the last few layers so we can classify images according to our desirable classes. This helps us get good results even with a small dataset since the basic image features have already been learnt in the pre-trained model from a much larger dataset

Continue…

Transfer Learining

Traditional Learning and Transfer Learning

Continue…  Inception V3: Inception-v3 is a convolutional neural network that is 48 layers deep. You can load a pretrained version of the network trained on more than a million images from the ImageNet database

Architecture of Inception V3

Continue…  Flutter : Flutter is an open-source UI software development kit developed by Google.

 Flutter is the tool that allows you to build native cross-platform(Android and IOS) Applications with one programming language and code base.  Flutter uses Programming language “Dart”.

 Dart is a Object oriented programming language.  It is a programming language which is focused on frontend(Android and IOS application) and User interface(UI) development.  This Dart programming language developed by google.

Continue…

Flutter/Dart transformed into native apps

Sample Code  Languages :- Python, Dart  Building a model

Continue…  Front end Mobile application

Continue…

Output

Conclusion Still the concept of medication and details about those disease are yet to be included. Even though this project already implemented, using Transfer Learning concepts, the accuracy of the result may increase and being a precompiled model the weights are predefined hence the time taken to build the model will decrease. This application is very useful for the farmers to detect the disease within no time, such that the prevention of the disease from a wide spread can be stopped.



Thank You